You are on page 1of 372

R=19710021465 2017-12-17T09:14:27+00:00Z



Volume Ill

Radiation Transfer With Absorbing,

Emitting, and Scattering Media


Volume III
Radiation Transfer With Absorbing,
Emitting, and Scattering Media

Robert Siegel and John R. Howell

Lewis Research Center
Cleveland, Ohio

Scientific and Technical I?~formationOffice 1971

Washington, D.C.
For sale by the Superintendent of Documents,
U.S. Government Printing Office, Washington, D.C. 20402
Price $1.75
Library of Congress Catalog Card Number 67-62877
This is the third and final volume of the series "Thermal Radiation
Heat Transfer" that is being published as the NASA Special Publication
SP-164. The first and second volumes appeared in 1968 and 1969,
respectively. As stated in the Preface to volume I, this publication is
an outgrowth of a course in thermal radiation that has been given as
part of an internal advanced study program at the NASA Lewis Research
This volume contains nine chapters. The first discusses some of the
fundamentals of absorption of radiation along a path in amedium, and
emission by the medium. The important property of the radiation
intensity is developed showing that along a path in a nonattenuating
nonemitting medium the intensity is invariant with position. Thus the
magnitudes of any attenuation or emission can be expressed in terms
of the changes produced in the intensity with distance.
In chapter 2 the absorption, scattering, and emission effects along a
path are combined to develop the equation of transfer. The solution
of this equation gives the intensity within the medium. The intensity
can then be used to obtain energy fluxes. These are related to the local
temperature within the medium by means of the energy conservation
equation. This provides the necessary relations to obtain energy transfers
and temperature distributions in the medium.
Some methods for solving these equations are given in chapter 3. These
methods include various approximate techniques, a very important one
being the diffusion approximation. The boundary conditions and
application of the diffusion method are considered in detail.
Before any radiation solution can be applied, the radiative properties
must be known. Gas radiation properties vary eonsiderably with wave-
length. In chapter 4 some of the fundamentals of the radiative properties
are discussed. These include line broadening mechanisms and band
absorption correlations.
Chapter 5 considers the important situation of radiation from a gas
that is well mixed so that it is isothermal. The method of analyzing heat
transfer in a gas filled enclosure is developed by extending the net
radiation method presented in volume I1 for enclosures containing a
nonabsorbing medium. The very useful concept of radiative mean beam
length is derived for use in gas energy exchange.

Chapters 6, 7, and 8 provide further development by discussing the

Monte Carlo method, radiation combined with conduction and convec-
tion, and scattering. Finally, chapter 9 treats a number of specialized
topics such as radiation in media with nonunity refractive index, and
radiation from flames.
MEDIA ........................................................................ 1
1.1 INTRODUCTION .......................................................... I
1.2 SYMBOLS .................................................................... 3
SION ....................................................................... 4
TION INTENSITY ...................................................... 8
1.5 THE ATTENUATION OF ENERGY .................................. 12
1.5.1 The Extinction Coefficient ............................................ 13
1.5.2 Radiation Mean Penetration Distance ............................. 14
1.5.3 Optical Thickness ....................................................... 14
1.5.4 The Absorption Coefficient........................................... 15
1.5.5TrueAbsorptionCoefficient.......................................... 19
1.5.6 The Scattering of Energy ............................................. 20
1.6 THE EMISSION O F ENERGY ......................................... 21
ERTIES ..................................................................... 24
1.7.1 Absorptance ............................................................. 24
1.7.2 Emittance ................................................................. 25
1.7.3 Transmittance............................................................ 29
RIUM ....................................................................... 32
1.9 CONCLUDING REMARKS ............................................. 34
REFERENCES ..................................................................... 34
SORBING-EMITTING GAS ....................................
2.1 INTRODUCTION ..........................................................
2.2 SYMBOLS ....................................................................
2.3 THE EQUATION O F TRANSFER ....................................
2.3.1 Derivation .................................................................
2.3.2 Integration by Use of Integrating Factor ..........................
2.4.1 Radiative Equilibrium ..................................................
2.4.2 Some Mean Absorption Coefficients ...............................
2.6 THE GRAY GAS ...........................................................
2.6.1 Transfer Equations .....................................................
2.6.2 Plane Layer Between Black Plates .................................
2.6.3 Use of Exponential Integral Functions ............................
2.8 CONCLUDING REMARKS .............................................
REFERENCES .....................................................................

TION OF T M N S F E R ................................................
3.1 INTRODUCTION ..........................................................
3.2 SYMBOLS ....................................................................
IN THE EQUATION O F TRANSFER .............................
3.3.1 The I'ransparent b a s Approximation ..............................
3.3.2 The Emission Approxin~ation........................................
3.3.3 The Cold Medium Approximation ..................................
3.4.1 Simplified Derivation of the Diffusion Equation ................
3.4.2 The General Radiation Diffusion Equation ....................... The Rosseland equation for local radiative flux............ The emissive power jump as a boundary condition ......... The emissive power jump between two absorbing-emit-
ting regions ........................................................ Summary.............................................................
3.4.3 Use of the Diffusion Solution ........................................ Gray stagnant gas between parallel plates .................. The discontinuity in emissive power between two gas
reglons .............................................................. Other diffusion solutions for gray gases .....................
3.4.4 Final Remarks ...........................................................
EFFICIENTS ..............................................................
3.5.1 Some Mean Absorption Coefficients ...............................
3.5.2 Approximate Solutions of the Transfer Equations Using
Mean Absorption Coefficients .....................................
TION O F TRANSFER .................................................
3.6.1 The Astrophysical Approximations ................................. The Schuster-Schwarzschild approximation ............... The Milne-Eddington approximation ..........................
3.6.2 The Differential Approximation ..................................... Houndary conditions ............................................. Applications of the differential approximation .............
3.7 CONCLUDING REMARKS .............................................
REFERENCES .....................................................................


PROPERTIES ............................................................ 113
4.1 INTRODUCTION ......................................................... 113
4.2 SYMBOLS .................................................................. 113
4.3 SOME ELEMENTS O F QUANTUM THEORY .................. 115
4.3.1 Bohr Model of Hydrogen Atom ...................................... 115
4.3.2 Schriidinger Wave Equation ....................................... 117
TION ......................................................................... 121

4.5 THE EQUATION OF TRANSFER ................................... 125
4.6.1 Spectral Line Broadening ........................................ 128 Natural broadening ................................................ 131 ............................................... 131 ............................................ 132 ............................................... 134
4.6.2 Absorption or Emission by a Spectral Line ...................... 134
4.6.3 Continuum Absorption ................................................ 137 Bound-free processes ............................................. 137 Free-free processes ............................................... 138
4.6.4 Band Absorption Correlations ................................. 138
4.7 CONCLUDING REMARKS ............................................. 153
REFERENCES ..................................................................... 153
RADIATION IN ENCLOSURES ............................. 157
5.1 INTRODUCTION .......................................................... 157
5.2 SYMBOLS ................................................................. 157
5.3.1 Definitions of Spectral Transmission and Absorption
Factors .................................................................. 164
5.3.2 Matrix of Enclosure Theory Equations ........................... 164
5.3.3 Heat Balance on Gas ................................................... 166
5.4.1 Hemisphere to Differential Area at Center of Its Base ......... 168
5.4.2 Top of Right Circular Cylinder to Center of Its Base ............ 169
5.4.3 Side of Cylinder to Center of Its Base ............................ 171
5.4.4 Entire Sphere to Any Element on Its Surface or to Its Entire
Surface .................................................................. 172
5.4.5 Infinite Plate to Area on Parallel Plate ........................... 173
5.4.6 Rectangle to a Directly Opposed Parallel Rectangle .......... 174
BOUNDARY ............................................................ 175
5.5.1 Mean Beam Length for Gas Between Parallel Plates Radi-
ating to Area on Plate ............................................... 177
5.5.2 Mean Beam Length for Sphere of Gas Radiating to Ally
Area on Boundary ................................................... 177
5.5.3 Radiation from Entire Gas Volume to Its Entire Boundary
in Limit When Gas Is Optically Thin ........................... 178
5.5.4 Correction for Mean Beam Length When Gas Is Not Op-
tically Thin ............................................................ 179
5.6.1 Radiation from Gas to All or Portion of Boundary ............... 183
5.6.2 Exchange Between Entire Gas Volume and Boundary ......... 187

5.7.1 Band Equations ......................................................... 189
5.7.2 Transmission and Absorption Factors ............................. 189
5.8.1 The Curtis-Godson Approximation ................................ 202
5.8.2 The Zoning Method ..................................................... 207
5.8.3 The Exchange Factor Approximation .............................. 211
REFERENCES ..................................................................... 219
6 T H E MONTE CARLO TECHNIQUE ....................... 221
6.1 INTRODUCTION .......................................................... 221
6.2 SYMBOLS .................................................................... 221
6.3 DISCUSSION OF THE METHOD ................................... 223
6.4 RADIATION THROUGH GRAY GASES .......................... 229
6.4.1 Infinite Parallel Planes ................................................ 229
6.4.2 Infinitely Long Concentric Cylinders .............................. 231
6.4.3 Radiation Between Adjacent Gray Keg~ons..................... 233
TIONS ....................................................................... 234
6.7 TRANSIENT RADIATION PROBLEMS ............................ 237
6.9 CONCLUDING REMARKS .............................................. 238
REFERENCES ...................................................................... 238
VECTION ...................................................................... 241
7.1 INTRODUCTION .......................................................... 241
7.2 SYMBOLS .................................................................... 241
7.3 RADIATION WITH CONDUCTION ................................ 243
7.3.1 The Conduction-Radiation Parameter ............................. 244
7.3.2 Energy Balance ........................................................... 246
7.3.3 Plane Layer Geometry .................................................. 247
7.3.4 Simple Addition of Radiation and Conduction Energy
Transfers ............................................................... 251
7.3.5 The Diffusion Method ................................................... 254
7.3.6 The Exchange Factor Approximation ............................... 260
7.4.1 Boundary Layer Problems ............................................. 265 Optically thin thermal layer ...................................... 266 Optically thick thermal layer .................................... 268
7.4.2 Channel Flows ............................................................ 270
7.4.3 Other Multimode Problems ....................................... 272
7.5 CONCLUDING REMARKS .............................................. 272
REFERENCES ...................................................................... 273

ABSORBING MEDIA ................................................
8.1 INTRODUCTION ..........................................................
8.2 SYMBOLS ....................................................................
TION O F SCATTERING .............................................
8.3.1 The Scattering Cross Section .........................................
8.3.2 The Phase Function .....................................................
8.4.1 A Cloud of Large Specularly Reflecting Spheres ...............
8.4.2 Reflection From a Diffuse Sphere ....................................
8.4.3 Large Dielectric Sphere with Refractive Index Close to
Unity ......................................................................
8.4.4 Diffraction From a Large Sphere ...................................
8.4.5 Rayleigh Scattering ...................................................... Scattering cross sections for Rayleigh scattering......... Phase function for Rayleigh scattering ........................
8.4.6 Mie Scattering Theory .................................................
8.5.1 The Equation of Transfer in a Pure Scattering Atmosphere ...
8.5.2 Scattering in Absorbing-Emitting Media ..........................
REFERENCES .....................................................................
ING-RADIATING MEDIA. .............................
9.1 INTRODUCTION ..........................................................
9.2 SYMBOLS ....................................................................
REFRACTIVE INDEX .................................................
9.3.1 Media With Constant But Nonunity Refractive Index ..........
9.3.2 The Effect of Brewster's Angle .......................................
9.3.3 Radiative Transfer Between Dielectrics Spaced Closely
Together .................................................................
TION ........................................................................
9.4.1 Theoretical Flame Temperature ......................................
9.4.2 Radiation From Nonluminous Flames ..............................
9.4.3 Radiation From and Through Luminous Flames ................. Experimental correlation of soot spectral absorption ...... Electromagnetic theory prediction of soot spectral
absorption .......................................................... Total emittance of soot cloud ....................................
9.4.4 Radiation From Gases Containing Luminous Particles .........
9.5 LUMINESCENCE ..........................................................
9.6 TRANSIENT RADIATION PROBLEMS ...........................
REFERENCES .....................................................................
APPENDIX .................................................. .. ...................................
INDEX ................................................................................................
Chapter 1. Fundamentals of Radiation in Absorbing,
Emitting, and Scattering Media
The study of energy transfer through media that can absorb, emit,
and scatter radiation has received increased attention in the past several
years. This interest stems from the complicated and interesting phe-
nomena associated with nuclear explosions, hypersonic shock layers,
rocket propulsion, plasma generators for nuclear fusion, and ablating
Although some of these applications are quite recent, the study of
gas radiation has been of continuing interest for over 100 years. One
of the early considerations was the absorption of radiation in the Earth's
atmosphere. This has always plagued astronomers when observing on
Earth the light from the Sun and more distant stars. Figure 1-1 shows
the observed form of the solar spectrum recorded by Samuel Langley
over a period of years beginning in 1880. The dashed curve shows the
estimated solar emission curve (assumed to be a blackbody spectrum
at 5600 K) and the solid curve shows the spectrum after atmospheric
attenuation (ref. 1). The absorption occurs in specific wavelength regions,
illustrating that gas radiation properties vary considerably with wave-
length. The atmospheric absorption of solar radiation is chiefly by water
vapor and carbon dioxide. Extensive discussions of absorption in the
atmosphere are given by Goody (ref. 2) and Kondratyev (ref. 3).
Gas radiation has also been of interest to astrophysicists with regard

Solar spectrum measured o n Earth

---- Solar emission spectrum


1.1 1.4 1.8 2.6 4.4
Wavelength, A, pin

FIGURE 1-1. -Attelluatio~l of solar spectrum by Earth's atmosphere as measured by Langley

(ref. I).

to the study of stellar structure. Models of stellar atmospheres, such

as for the Sun, and the energy transfer processes within them have been
constructed; then the emitted energy spectra calculated on the basis
of the models are compared with observed stellar spectra.
In industry the importance of gas radiation was recognized in the 1920's
in connection with heat transfer inside furnaces. The carbon dioxide
and water vapor formed as products of combustion were found to be
significant emitters and absorbers of radiant energy. Radiation can also
be appreciable in engine combustion chambers where peak temperatures
reach a few thousand degrees. The energy emitted from flames depends
not only on the gaseous emission but also arises from the heated carbon
(soot) particles that are formed within the flame.
Another interesting example of radiation within an absorbing-emitting
medium is in a glass melting furnace. As described in reference 4,
the temperature distribution measured within a deep tank of molten
glass was found to be more uniform than that expected from heat
conduction alone. It was thought that convection might account for the
discrepancy, but experimental investigations did not indicate that this
was the contributing heat transfer mode. In the late 1940's it became
evident that radiative transfer by absorption and reemission within
the glass provided a significant means of energy transport.
Two difficulties are encountered in studying radiation within absorbing,
emitting, and scattering media that make these studies, to say the least,
challenging. First, absorption and emission of energy are occurring not
only at system boundaries, but at every local point within the medium.
Scattering is also a local transfer process within the medium. A complete
solution of the energy exchange problem therefore requires knowledge
of the temperature and physical properties of the medium at every point
within the system. The mathematics describing such a situation is in-
herently complex. A second difficulty is that spectral effects are often
much more pronounced in gases than for solid surfaces. As a result a
detailed spectrally dependent analysis may be required. When approxi-
mations are used based on spectrally averaged properties, special care
must be taken. Most of the simplifications introduced in gas radiation
problems are aimed at dealing with one or both of these two complexities.
A brief comment on the approach to radiation in gases used herein is
in order. The astrophysical approach (i.e., the equation of transfer) is
used for determining the local values of the radiation intensity within the
medium. As will be defined in detail in section 1.4, the intensity is con-
cerned with energy transport along a path in a single direction. By study-
ing the variation of intensity along a path, a good understanding is
obtained of how the individual processes of absorption, emission, and
scattering enter into the radiative transfer. It is the most useful method

in problems dealing with atmospheric absorption, stellar structure, and

others where the spectral intensity at some position is often a quantity
of interest. Two excellent texts (refs. 5 and 6) deal in detail with this
formulation as used in astrophysics.
The astrophysical approach however must be adapted for more con-
venient use by the engineer. The engineer is chiefly interested in energy
fluxes and temperatures rather than radiation intensities. Also the astro-
physical notation and nomenclature is foreign to most engineers and is
often inconsistent within itself. For these reasons, although fundamental
ideas are developed here on the basis of the intensity of the radiation,
the change to terms of local energy flux and temperature is often made.
This change aids in developing useful engineering solution methods and
will also show how the engineering methods can be derived in a logical
manner from the astrophysical relations.

absorption coefficient
concentration of gas in mixture
second constant in Planck's spectral energy distribution,
speed of light in medium other than vacuum
speed of light in vacuum
ionization potential
e emissive power
Fo-kT fraction of blackbody emissive power in spectral region
h, Planck's constant
i radiation intensity
K +
extinction coefficient, a (TS
k Boltzmann constant
111, extinction mean free path
n simple refractive index
P pressure
P partial pressure of gas in mixture
Q energy per unit time
4 energy flux, energy per unit area and time
R radius of sphere
S coordinate along path of radiation
T absolute temperature
V volume
CY absorptance

cone angle, angle from normal of area

wave number
optical thickness (eq. (1-17))
wavelength in medium
Stefan-Boltzmann constant
scattering coefficient
solid angle

a absorbed
b blackbody
e emitted
g gas
i component i
m mass coefficient; mean value
P projected
s source or scatter
"I wave number dependent
h wavelength dependent
I/ frequency dependent

directional quantity
+* true value, not modified by addition of induced emission
dummy variable of integration


Although this volume will be concerned with radiation in absorbing,
emitting, and scattering media in general, it will almost always be gases
that are used a s examples. If the radiation properties of gases and opaque
solids are compared, a difference in spectral behavior is quite evident.
As shown by the plots of radiation properties in chapter 5 of reference 7
(which will be referred to from this point as vol. I), the property variations
with wavelength for opaque solids are fairly smooth although in some
instances the variation is somewhat irregular. Gas properties however
exhibit very irregular wavelength dependencies. As a result the absorp-
tion or emission by gases is significant only in certain wavelength regions,
especially at temperature levels below a few thousand degrees Kelvin.


Wavelength, A, p m

Wave number, 7,c m - I

1-2.-Low-resolution spectrum of absorption bands for COz gas at 830 K, 10 atm,
and for path length through gas of 38.8 c n ~ .

The absorptance of a gas layer as a function of wavelength typically

looks as shown for carbon dioxide (COz)in figure 1-2.
The radiation emitted from a solid actually originates within the solid
so the solid can be considered an absorbing and emitting medium like
a gas; the physics of the radiation thus has a common basis for all media.
The differences in spectra are caused by the various types of energy
transitions that occur within the media. A gas has different types of
transitions, a fact which leads to a less continuous spectrum than for a
solid. The energy transitions that account for radiation emission and
absorption will now be discussed.
A radiating gas can be composed of molecules, atoms, ions, and free
electrons. These particles can have various energy levels associated
with them. In a molecule, for example, the atoms form a dynamic system
that has certain vibrational and rotational modes. These modes have
specific energy levels associated with them. A schematic diagram of
the energy levels for an atom, ion, or electron is shown in figure 1-3.
(The levels for a molecule are diagramed in fig. 4-5.) The zero energy
level is assigned to the ground state (lowest energy bound state) with
the higher bound states being at positive energy levels. The energy E I



(a), Bound-bound absorption

(b), Bound-bound emission
(c), Bound-free absorption
(d), Free-bound emission
(e), Free-free absorption
(f), Free-free emission

1-3. -Schematic diagram of energy states and transitions for atom, ion, or electron.

in figure 1-3 is the ionization potential, that is, the energy required to
produce ionization from the ground state. Energies above El denote that
ionization has taken place and free electrons have been produced.
It will be convenient to discuss the radiation process by utilizing a
photon or quantum point of view. The photon is the basic unit of radiative
energy. Radiative emission will consist of the release of photons of
energy, and absorption will be the capture of photons by a particle.
When a photon is emitted or absorbed, the energy of the emitting
or absorbing particle will be correspondingly decreased or increased.
Figure 1-3 is a diagram of the three types of transitions that can occur.
These are bound-bound, bound-free, and free-free; they will be discussed
a little further on in more detail. In addition to emission and absorption
processes, it is possible for a photon to transfer part of its energy in

certain inelastic scattering processes. These are of minor importance in

engineering radiative transfer.
The magnitude of the energy transition is related to the frequency of
the emitted or absorbed radiation. The energy of a photon is hv where
h is Planck's constant and v is the frequency of the photon energy.
For an energy transition, say from bound state E3 down to bound state E2
in figure 1-3, a photon is emitted with energy E 3-E B= hv. The frequency
of the emitted energy is then v = (E3- E 2 ) / h so that a fixed frequency
is associated with the transition from a specific energy level to another.
Thus in the absence of any other effects, the radiation emitted will be
in the form of a spectral line. Conversely in a transition between two
bound states when a particle absorbs energy, the quantum nature of
the process dictates that the absorption is such that the particle can
only go to one of the discrete higher energy levels. Consequently, the
frequency of the photon energy must have certain discrete values in
order for the photon to be absorbed. For example, a particle in the ground
state in figure 1-3 may absorb photons with frequencies (EZ-E,)/h,
(E3 -EJh, or (E4 - El)/h and undergo a transition to a higher bound
energy level. Photons with other frequencies in the range 0 < v < E,/h
cannot be absorbed.
When a photon is absorbed or emitted by an atom or molecule and
there is no ionization or recombination of ions and electrons, the process
is termed a bound-bound absorption or emission (see processes (a)
and (b) in fig. 1-3). The atom or molecule moves from one quantized
bound energy state to another. These states can be rotational, vibra-
tional, or electronic in molecules, and electronic in atoms. Since the
bound-bound energy changes are associated with specific energy levels,
the absorption and emission coefficients will be sharply peaked functions
of frequency in the form of a series of spectral lines. These lines do have
a finite width from various broadening effects that will be discussed in
section 4.6.1.
The vibrational energy modes are always coupled with rotational
modes. The -rotational spectral lines superimposed on the vibrational
line give a band of closely spaced spectral lines. If these are averaged
together into one continuous region, it becomes a vibration-rotation
band (see section 4.6.4). Rotational transitions within a given vibrational
state are associated with energies at long wavelengths, 8 to 1000 pm
(see fig. 1-2 of vol. I). Vibration-rotation transitions are at infrared ener-
gies of about 1.5 to 20 pm. Electronic transitions are at short wavelengths
in the visible region, 0.4 to 0.7 pm, and at portions of the ultraviolet and
infrared near the visible region. At industrial temperatures the radiation
is principally from vibrational and rotational transitions; at high tempera-

2.4.1 of vol I). It is as if the radiation traveling through an area within

the medium originated at that area. The intensity is then deJ;ned (see
fig. 1-4(a)) as the radiation energy passing through the area per unit
time, per unit of the projected area andper unit solid angle. The projected
area is formed by taking the area that the energy is passing through and
projecting it normal to the direction of travel. The unit elemental solid
angle is centered about the direction of travel and has its origin at dA.
The spectral intensity is the intensity per unit small wavelength interval
around a wavelength A.
As stated previously, the emitted intensity from a blackbody is
invariant with emission angle. Now a second invariant property of
intensity will be examined. Consider radiation from a source d A , travel-
ing in an ideal medium that is nonabsorbing, nonemitting, and non-
scattering and has constant properties. Suppose that an imaginary area
element ~ A isI considered at distance S 1 from dA, and that dAl is normal
to S1 as shown in figure 1-4(b). From the definition of spectral intensity
i i , l a s the rate of energy passing through d A l per unit projected area
of d A l per unit solid angle and per unit wavelength interval, the energy
from d A , passing through dA1 in the direction of SI is

where the third derivative notation d3 emphasizes that there are three
differential quantities on the right side of the equation. The solid angle
d w ~is equal to dA,/S: so

Suppose that d A l is now placed a distance Sn from the source along the
same direction a s for the original position. The rate of energy passing
through dAl in the new position is

Dividing equation (1-lb) by equation (1-2) gives

Now consider a differential source emitting energy equally in all


Projection of dA into plane

Radiation from .~ e r. ~ e n d i c u l toa r intensitv.
direction p and
w i t h i n solid angl
dw that passes
t h r o u g h dA
Direction of
travel of intensity

(a) Geometry for definition of intensity in medium.

(b) Intensity from source to area element.
( c ) Variation of energy flux with (d) Intensity of emitted uadiation.
distance from source.

1-4.-Derivations of intensity relations.

directions and draw two concentric spheres around it as in figure 1-4(c).

If CPQA,~is the entire spectral energy leaving the source, then the energy
flux crossing the inner sphere is d 2 Q h , , y / 4 ~ Sand
: that crossing the
outer sphere is d 2 Q h , , / 4 ~ S $The
. ratio of the energies passing through
the two elements dAl is

Substituting equation (1-4) for the left side of equation (1-3) gives the
following important result:

Thus, the intensity i n a given direction i n a nonattenuating and non-

emitting medium with constant properties is independent of position
along that direction. Note that these intensities are based on the solid
angles subtended by the source as viewed from d A 1 as in figure 1-4(b).
As S is increased, the decrease in solid angle by which d A l views the
source d A , is accompanied by a comparable decrease in energy flux
arriving at d l .Thus the flux per unit solid angle, used in forming the
intensity, remains constant.
The radiant energy passing through dAl can also be written in terms
of the intensity leaving the source. Using figure 1-4(d) results in

Equating this with the energy rate passing through dAl a s given by
equation (1-lb) results in
i);,l = ",s (1-7)

This relation again shows the invaliance of intensity with position in a

nonattenuating and nonemitting medium.
The invariance of intensity when no attenuation or emission is present
provides a convenient way of specifying the magnitudes of any attenua-
tion or emission as these effects are given directly by the change of
intensity with distance. By use of the foregoing intensity properties, the
attenuation and emission of radiation within a medium can now be


Consider spectral radiation of intensity i i impinging normally on a

layer of material of thickness dS as in figure 1-5. The medium in the
layer absorbs and scatters radiation. For the present it will be assumed
that the layer is at low temperature so that its emitted energy is negligible.
As the radiation passes through the layer, its intensity is reduced by
absorption and scattering. The change in intensity has been found
experimentally to depend on the magnitude of the local intensity. If a
coefficient of proportionality I<A which depends on the local properties
of the medium is introduced, then the decrease is given by

This equation contains the assumption that no intensity is scattered

from the radiation field into the direction of S.
The quantity I(Ais called the extinction coeficient of the material in
the layer. The extinction coefficient is a physical property of the material
and has the units of reciprocal length. It is in general a function of the
temperature T, pressure P , composition of the material (specified here
in terms of the concentration Ci of the i components), and the wavelength
of the incident radiation so that

Kh= Ki(X, T , P , Ci) (1-9)

As will be shown later (see eq. (1-16)), the KAis inversely related to the
mean penetration distance of radiation in an absorbing and scattering
Integrating equation (1-8) over a path of length S gives the relation

1-5. -Intensity incident normally on absorbillg and scattering layer of thickness dS.

where i i ( 0 ) is the intensity entering the layer and S" is a dummy vari-
able of integration. Integrating equation (1-10) yields

K A( S * )dS*

i i ( S )= i i ( 0 ) exp [- IOs
KA(s*)~s*] (1-12)

Equation (1-12) is known as Bouguer's law it shows that, a s a conse- ';

quence of the proportionality in equation (1-8), the intensity of mono-
chromatic radiation along a path is attenuated exponentially while
passing through an absorbing-scattering medium. The exponent is equal
to the integral of the local extinction coefficient over the path length
traversed by the radiation.

1.5.1 The Extinction Coefficient

The extinction coefficient for thermal radiation I<A is composed of
two parts, an absorption c o e f i i e n t aA(X, T, P ) and a scattering coe&ient
r S A ( X , T, P ) . For simplicity the notation has been dropped showing de-
pendence upon the relative concentration of the constituents of the gas.
The coefficients are related by

As noted previously, these coefficients have units of reciprocal length

and are therefore called linear coefiients. Some researchers prefer
to work with mass coefficients given by

KA,,I,= ah, n t + r d , m=-=-

K A ah
+-P ~ S A

where p is the local density of the absorbing-scattering species. The

mass coefficients have units of area per unit mass and are directly related
to the concept of a cross section in molecular physics (see section 8.3.1
for a discussion of scattering cross sections). Since the extinction
coefficient K A increases as the density of the absorbing or scattering

' Named after Pierre Bouguer (1698 to 1758)lwho first showed on a quantitative basis haw light intensities could b e
compared. Equation (1-12) is sometimes called Lambert's law, the Bouguer-Lambert law, or Beer's law. Beer's law is
more properly a restricted form of equation (1-9) stating that the absorption of radiation depends only on the concen-
tration of t h e absorbing species along the path. l o avoid confusion rvith Lambert's cosine law, equation (1-12) will b e re-
ferred to herein as Bauguer's law.

species is increased, the use of Kx,,,,=KA/p has the advantage that it

tends to remain more constant than KA. However, the I<* which will be
used in this volume also has an advantage in that, when I < A is constant,
it can be interpreted as the reciprocal of the radiation mean penetration
distance. This will now be shown.

1.5.2 Radiation Mean Penetration Distance

From equation (1-12) the fraction of the original radiation that travels
through the path length S is

ii(S) - exp
[- loS

The fraction absorbed in the layer from S to S +dS is

The mean penetration distance of the radiation is obtained by multi-

plying the fraction absorbed at S by the distance S and then integrating
over all path lengths from S = 0 to S = m; that is,

When KAis constant, carrying out the integral gives

11. = KA 1' 1
S exp (- KAS)d s = - (1-16)

demonstrating that the average penetration distance before absorption

or scattering is the reciprocal of KA when I<A does not vary along the
path. Equation (1-16) provides a simple way of gaining some insight as
to whether or not an absorbing-scattering medium is very opaque with
regard to radiation traveling through it. This will now be further dis-
cussed in connection with the definition of optical thickness.

1.5.3 Optical Thickness

The exponential factor in equation (1-12) is often written in an alternate
form by defining

so that equation (1-12) becomes

i ; ( S ) = i i ( 0 ) exp [- K A ( S1) (1-18)

The quantity K A ( S )is the optical thickness or opacity of the gas layer
of thickness S and is a function of all the values of KA that lie between
0 and S. Because KA is a function of the local parameters P, T, and Ci,
the optical thickness becomes a function of all these conditions along
the path between 0 and S.2
The optical thickness is a measure of the ability of a given path
length of gas to attenuate radiation of a given wavelength. A large
optical thickness means large attenuation. The quantity K A is a con-
venient dimensionless parameter that will occur in the solutions of
radiative transfer problems.
For a gas that is of uniform composition and is at uniform temperature
and pressure (a un$orm gas) or for a gas with KA independent of T, P,
and Ci, equation (1-17) becomes

K A ( S ) = I(k.5 (1-19)

The optical thickness then depends directly on the extinction coefficient

and the thickness of the absorbing-scattering layer.

1.5.4 The Absorption Coefficient

If scattering can be neglected (i.e., aSx= 0), then I<A= aA and equation
(1-12) becomes

i ; ( S ) = i;(O) exp [- [~ * ( s * ) B * ] (1-20)

If, in addition, ax is not a function of position as is the case in a gas of

uniform temperature, pressure, and composition, then

i i ( S )= iL(0) exp (- U A S ) (1-21)

'The notation for the optical thickness ti* should not he confused with the extinction coefficient for electromagnetic
radiation ti used in equations (1-22) and (1-23) that follow. It is regrettable but true that the notation possibilities of the
English and Greek alphabets reach saturation when such interdisciplinary fields as gas radiation are discussed.

In the electromagnetic theory of the propagation of radiant energy

(see discussion following eq. (4-26) in vol. I), it is shown that the inten-
sity of radiation is attenuated in conducting media according to the

where K is the extinction coefficient from electromagnetic theory and is

related to the magnetic permeability, electrical resistivity, and electrical
permittivity of the medium (eq. (4-23b) in vol. I). Thus, ah is related to
K by

Such a relation provides some theoretical basis for Bouguer's law, which
was originally based on experimental observations.
The absorption coefficient a ~ ( h T,
, P) usually has strong variations
with wavelength and often varies substantially with temperature and
pressure. Considerable analytical and experimental effort has been ex-
pended in the determination of a , for
~ individual gases.
Analytical determinations of ah require detailed quantum mechanical
calculations beyond the scope of this volume, although some of the
concepts of these calculations are outlined in chapter 4. Except for the
simplest gases such as atomic hydrogen, the calculations are very tedious
and require many simplifying assumptions. For the methods used in
calculation of C I A , references 8 to 10 give detailed discussions.
The complexity of the calculations is presaged by examination of some
measured solid and gas spectral absorption coefficients. In figure 1--6,
ah is shown for pure diamond. Strong absorption peaks due to crystal
lattice vibrations at certain wavelengths are evident. Figure 1-7 shows
the calculated emission spectrum of hydrogen gas at 40 atm and 11300 K
for a path length through the gas of 50 cm. The variations in emission
are closely related to variations in the absorption coefficient. The
presence of "spikes" or strong emission lines is the result of transitions
between bound energy states. The continuous part of the emission
spectrum is due to various photodissociations, photoionization, and free
electron-atom-photon interactions of other types. The lines and the
continuous regions are common features of both emission and absorption
spectra. Figure 1-8 shows the absorption coefficient of air at 1 atm and
12 000 K. In this case, there is a merging of the contributions from the
many closely spaced lines produced by vibrational and rotational

1-6.-Spectral absorption coefficient of diamond (from ref. 15).

1111 Contributions from

discrete transitions
-- Blackbody emission


Wave number, v, cm-I

FIGURE1-7.-Normalized emission spectrum of hydrogen at 11 300 K, 40 atnl, and for

path length of 50 cm (from ref. 16).

Frequency, v, sec''

FIGURE1-8.-Absorption coefficient of air at 12 000 K and 1 atm (from ref. 17).

transitions between energy states, and the absorption coefficient has

the appearance of being continuous. Even when this merging is not
complete, the resolution of experimental measurements causes the
measured spectrum to appear continuous over these closely spaced
It is noted that figures 1 4 , l - 7 , and 1-8 each have a different abscissa
(i.e., wavelength, wave number, and frequency), emphasizing the lack
of an accepted standard variable. When radiative properties of opaque
surfaces were discussed in volume I of this series, it was found that
the wavelength was generally used. In radiation from gases, however,
the frequency is more common. It has the advantage that the frequency
does not change when radiation passes from one medium into another
with a different refractive index. The wavelength does change because
of the change in propagation velocity.
Dealing with spectral line emission and absorption is one of the
computational difficulties encountered in analyses of radiant energy
transfer through gases. Incident radiation at wavelengths near the line
center will be strongly absorbed, while radiation of only a slightly
different wavelength may experience almost no attenuation. Integrating
line absorption coefficients with respect to wavelength to obtain band
or total absorption coefficients is generally tedious. These averaged
coefficients are used in certain calculation methods of radiative transfer.

1.5.5 True Absorption Coefficient

Bouguer's law in the form of equation (1-20) gives the attenuation of a
beam of radiation upon passing through a volume of nonemitting non-
scattering gas along a path of length S as would be observed by detectors
of incident and emerging radiation. Such observed information could be
used in determining a h . Actually, as radiative energy passes through a
gas, not only is it absorbed but there is an additional phenomenon in that
its presence stimulates some of the gas atoms or molecules to emit
energy. This is not the ordinary or spontaneous emission that will be
discussed in section 1.6. The spontaneous emission is the result of the
excited state of the gas being unstable and decaying spontaneously to a
state of lower energy. The emission resulting from the presence of the
radiation field is termed stimulated or induced emission and is in a sense
a negative absorption.
Physically, the induced emission process can be pictured a s follows:
A photon of a certain frequency from the radiation field encounters a
gas atom or molecule in an excited state, that is, an energy state above
the ground state. There exists a certain probability that the incident
photon will trigger a return of the gas particle to a lower energy state.
If this occurs, the particle will emit a photon at the same frequency and
in the same direction a s the incident photon. Thus, the incident photon
is not absorbed but is joined by a second identical photon. This process
is often viewed as a negative absorption and is so treated in the equations
of energy balance to be derived in chapter 2. More discussion of the
induced emission process is given in section 4.4.
The induced emission constitutes a portion of the intensity that is
observed in the beam emerging from the gas volume. Consequently,
the amount of energy that was actually absorbed by the gas is greater
than that found by taking the difference between the entering and leaving
intensities. This is because the observed emerging intensity is the result
of the actual absorption modified by the addition of induced emission
along the path of the beam. The actual absorbed energy should be
calculated using a true absorption coefficient ax+(A,T, P ) which will be
larger than the absorption coefficient ux(h, T, P ) calculated by using
observed attenuation data and Bouguer's law. The "true" law for
absorption of energy along path S is then written as

Statistical mechanical considerations give the relation between

cr~(X,T, P) and ax+(A, T, P ) for a gas with refractive index 17 = 1 as

.,(A, T , P ) = [1- exp (-$)I a t (A. T , P )

= [1 - exp (-s)] o$ (A, T, P ) (1-25)

Examination of equation (1-25) shows that, because of the negative

exponential term, a? will always be larger than ax (hence the use of the
superscript +).
Because the induced emission depends on the incident radiation field,
it is usually grouped together with the true absorption thereby yielding
the absorption coefficient ax. The emission term in the equation of
radiative transfer then includes only the spontaneous emission and
consequently depends only on the local conditions of the gas. As will
be shown in chapter 2, the grouping of induced emission into the ahsorp-
tion term simplifies the equations of radiative transfer.
The exponential term in equation (1-25) is small except at large values
of AT. Thus ah and ahf are nearly equal except at large values of AT (long
wavelengths and/or high temperatures). The values are within 1 percent
for XT less than 3120 (pm)(K) and within 5 percent for AT less than 4800
When properties from the literature are used in calculations of radia-
tive transfer in absorbing-emitting media, care must sometimes be
exercised to determine whether the reported absorption coefficients
include the effects of induced emission; usually it is ax that is given.

1.5.6 The Scattering of Energy

Scattering is taken here to be any encounter between a photon and
one or more other particles during which the photon does not lose its
entire energy. It may undergo a change in direction, and a partial loss or
a gain of energy. In any of these cases, the photon is said to have been
The scattering coeficient a s k is the inverse of the mean free path that
a photon of wavelength A will travel before undergoing scattering. (This
is strictly true only when a s h does not vary along the path.) The scatter-
ing can be characterized by four types of events: elastic scattering in
which the energy (and, therefore, frequency and wavelength) of the
photon is unchanged by the scattering, inela,stic scattering in which the
energy is changed, isotropic scattering in which scattering into any direc-
tion is equally likely, and anisotropic scattering in which there is a dis-
tribution of scattering directions. Elastic-isotropic scattering is most
amenable to analysis without resorting to sophisticated analytical or

numerical techniques. Most scattering events of importance in engineer-

ing are elastic, or very nearly so.
For an elastic scattering process, there is no exchange of energy be-
tween the radiation field and the medium. Therefore, the local thermo-
dynamic conditions of the gas are not affected by the radiation field,
although the radiation field is affected by the gas conditions. Scattering
calculations in this case become more tractable than for analogous
absorption-emission effects where the internal energy of the gas and
radiation field can interact strongly. Radiative transfer when scattering
is present is treated in chapter 8. Until that point, attention will be re-
stricted to cases involving only absorption and emission of radiation.


Having considered the various definitions connected with attenuation
within a medium, the emission of energy within the medium will now be
Consider an elemental volume dV of gas as shown in figure 1-9. The
true absorption coefficient within dV is ax+(A,T , P) and is considered
constant over dV. Let dV be placed at the center of a Large black hollow
sphere of radius R at uniform temperature T. The space between dV
and the sphere walls is filled with a nonparticipating material. The
spectral intensity incident at the dA, location on dV from an element
dA on the surface of the enclosure is, by use of equation (1-7),

FIGURE1-9.- Geometry for derivation of emission from volume of gas.


The intensity from this entering radiation that emerges from dV after
passing through the path dS is given by Bouguer's law as

i ; ( d S ) = i;(O) exp (- ah+&) (1-27)

Since the true absorption coefficient has been used, the i i ( d S ) from
equation (1-27) does not include any energy triggered by induced
The change of the intensity as a result of true absorption in dV is the
difference between equations (1-26) and (1-27); that is,

dii;=ii(dS)-ii(O)=--i;(O)[l-exp (-crxfdS)]
=- iib(A, T ) [ l - exp (- ax+dS)] ( 1-28)

This does not include induced emission. For very small values of ahfdS,
note that

di - zib(A, T ) [ l - exp (- a i d s ) ] - - i i b ( A ,T ) u i d S (1-29)

which will apply since dS is a differential quantity. Equation (1-29) is

also immediately evident from the differential form (eq. (1-8)). The energy
absorbed by the volume dSdA, from this incident radiation is

where dw = d A / R 2and dA, is a projected area normal to i i ( 0 ) . Substitut-

ing equation (1-29) in equation (1-30) results in

The energy emitted by dA and absorbed by all of dV is found by integra-

tion over d V ; that is,

d3Qi,, = Id1 d4Qi,, = iib( A , T )a i d h d o

projected c ~ r e f i
of d l . aor~~cctl
to p~it11Jro111dA
=atiLb ( A , T)dAdwdV (1-32)

where dw is the solid angle subtended by d A when viewed from dV.

To account for all energy incident upon dV from the entire spherical
enclosure, integration is carried out over all such solid a n d e s to give

~'QA,.= Jw d 3 ~ ; , , =a $ i [ b ( ~T)ilVdA

= 4.wa$iLb( A , T)dVdA

= 4a$e~b(A,
T) dVdA (1-33)

where exb is the blackbody spectral emissive power (eq. (2-12) in vol. I).
To maintain equilibrium in the enclosure, dV must emit an amount of
energy equal to that absorbed. Hence, the energy emitted by an iso-
thermal volume element in equilibrium with its surroundings is

This result includes both spontaneous emission and emission induced

by the incident equilibrium radiation field. For only spontaneous emis-
sion, the coefficient ah would be used. The shape of the element dV is
arbitrary; however, its size must he small enough to justify the approxi-
mation of equation (1-29) and also small enough so that energy emitted
within dV escapes before reabsorption within dV. Further, the gas must
be in thermodynamic equilibrium with respect to its internal energy, a
restriction discussed more fully in section 1.8.
At this point an emission coefficient could be defined in a similar
manner to the absorption coefficient. However, the radiation literature
has other definitions of the emission coefficient which do not follow an
analogy to the absorption coefficient definition, and there is no need to
add confusion by defining a new coefficient here. Rather, equation (1-34)
will be used directly as the relation for the emission of energy from an
infinitesimal volume element of gas.
When the spontaneously emitted intensity is the same for all directions
(isotropic spontaneous emission of energy), which is the condition for
all cases discussed in this volume, the radiation intensity emitted
spontaneously by a volume element into any direction is

"n the astrophysical literature (refs. 5 and 6), the emission coefficient is usually given the synlbol j~ defined by

having units, therefore, of energy rate per unit volume per unit wavelength interval.

where dAp is the projected area of dV normal to the direction of emission

and dS is the mean thickness of dV parallel to the direction of emission
(i.e., ci.5 = dV/dA,).


It is desirable to make use of the extensive techniques in the engineer-
ing literature dealing with radiative interchange computations between
surfaces (see ref. 11 (which will be referred to from this point a s vol. 11))
without intervening absorbing media. With this objective, analogous
concepts and terminology will now be developed for problems involving
participating gases. This is done through the concept of the emittance
and absorptance of a gas volume. These gas property definitions are
analogous to the emissivity and absorptivity of opaque bodies. Because
the energies emitted or absorbed by a volume of gas depend on the size
and shape of the volume in addition to its physical properties and
temperature, the absorptance and emittance are extensive properties.
The nomenclature used here will follow that of volume I1 of this series,
where the "ance" suffix applies for extensive properties.

1.7.1 Absorptance
To be a reasonably simple engineering parameter, the absorptance
should depend at most on the geometry, size, temperature, and physical
properties of the volume for which it is evaluated. It is, therefore, defined
for a volume with uniform conditions, so that no gradients in the physical
conditions need be considered.
Consider energy of intensity i i ( 0 ) incident in the S direction on a uni-
form medium of projected area d A , normal to S. The radiation passes
through a thickness S as shown in figure 1-10. The amount of energy in
solid angle dw absorbed by the medium is

d3Q):, = [i): (0) - i):( S )] dApdwdX (1-36)

LVolume of

FI(;URE1-10.-Geometry for absorptance along path length S .


By substituting equation (1-21) for the attenuation of intensity by a

uniform medium, this becomes

d"L,(! = i i ( 0 ) [ l - exp (- ahS) jdAl,dwdA (1-37)

The energy incident upon dV in solid angle dw is

If the absorptance for path length S in the volume is defined as the frac-
tion of the incident energy in solid angle dw that is absorbed while travers-
ing S in the volume, then dividing equation (1-37) by equation (1-38) gives

Spectral absorptance for path length S i n a un.$orm gas volunze

=ai(A, T , P, S )= 1'" = 1 - exp (-CLAS) (1-39)

By substituting equation (1-39) into equation (1-21) the relation between

intensities is obtained as

The a i ( S ) is a directional spectral absorptance. The values of absorp-

tance averaged over wavelength are also of use in engineering analyses.
Integrating equations (1-37) and (1-38) over all wavelengths and then
taking the ratio provides the relation

Total absorptance for path length S i n a uniform gas volunze

= a l ( T ,P-, S:) =
dlQ(1 -


i;(0) [ I - exp (-

U A S ) ] d~

1.7.2 Emittance
The directional emittance of a uniform gas volume is the ratio of the

energy emitted by the volume in a direction to that emitted by a black-

body at the same temperature. Because Kirchhoff's law holds without
restriction for directional spectral absorptance values, as discussed in
table 3-11 of volume I, it immediately follows by use of equation (1-39)

Directional spectral emittance for path length S i n a uniform gas volume

= €;(A, T , P , S ) = 1 - exp (- axS) (1-42)

The emittance is the ratio of emitted energy to that emitted by a black-

body. Consequently, the energy in wavelength interval d h arriving at
dA, at location S a s a result of emission by the medium in solid angle dw
as shown in figure 1-10 is i L b [ l- exp ( - a ~ S ) ] d A , d o d A . It follows by
analogy to equation (1-41) that for total quantities

Directional total emittance for path length S i n a uniform gas volume

- E' (T,P, S ) =
L x i i u ( h , T ) [ l - exp (- a h S ) ] d h

lox i;a(h, ~ ) d h

where by using c r T q o r the total blackbody emission it is assumed that

the index of refraction of the medium is n = 1.
In figure 1-11, the directional total emittance of carbon dioxide is
shown as a function of temperature, partial pressure of the COz, and path
length. This is an example of the extensive tabulations of such properties
that are available for gases at conditions of importance in industrial
design. The methods of using these properties in radiative exchange
computations are developed in chapter 5, where more detailed charts
of the radiative properties will be given.
Comparing equations (1-41) and (1-43) shows that Kirchhoffs law for
directional total properties, which is

---- From data


Temperature, "R

400 800 1200 1600 2000 2400 2800
Temperature, K

FIGURE1-11.-Emittance of carbon dioxide in mixture with nonabsorbing gas at total

pressure of 1 atm (ref. 18).

or' ( T , P , S ) = e r( T , P , S ) (1-44)

holds only under the conditions that the incident spectral radiation for
absorption is proportional to a blackbody spectrum at the gas tempera-
ture T , or the gas is gray, that is, ai = €1are independent of wavelength.
The same restrictions apply for opaque bodies as discussed in volume I.

EXAMPLE1-1: As a rough approximation, idealize the absorptance of

COz at T,= 1500" R (830 K ) and 10 atm as in figure 1-2 so that it consists
of four bands having vertical boundaries at the values 1.8 and 2.2, 2.6
and 2.8, 4.0 and 4.6, and 9 and 19 pm. What is the total emittance of a
very thick layer of gas at this temperature?
For a very thick layer of gas, equation (1-42) indicates that EL will go
to unity in the absorbing regions. Hence, the gas will emit like a black-
body in the four spectral absorption bands. In the nonabsorbing regions

between the bands, EL will be zero. From equation (1-43) the total
emittance becomes

E' (TfI, P , S) =
EL ( A . T,. P . S)e*u,,dA
- I
6niids -

The emittance is thus the fractional emission of a blackbody over the

wavelength intervals of the absorbing bands which can be obtained from
the FO-AT!~factors in table V in the appendix of volume I. The required
values are as follows:

Then the emittance is

Using the numerical values gives

EXAMPLE1-2: What fraction of incident solar radiation will be

absorbed by a very thick layer of COz at 10 atm and 1500" R (830 K)?
Use the approximate absorption bands of example 1-1.

The effective radiating temperature of the Sun is about T,$=10 000" R

(5600 K). The desired result is the fraction of the solar spectrum that lies
within the four COz bands as this is the only portion of the incident
radiation that will be absorbed. Using the FO-us factors as in example
1-1, but using the solar temperature, gives the following values (using
table V in the appendix of vol. I):

The fraction absorbed is then

Even though the gas layer is very thick, only 4.4 percent of the incident
energy is absorbed since the gas is essentially transparent in the region
between the absorption bands.

1.7.3 Transmittance
The transmittance of a gas volume is the fraction of the incident energy
that passes through the gas volume. If it is assumed that no reflection or
scattering of the incident radiation occurs, then the energy transmitted
along a path is the incident energy minus the energy absorbed along the
path: that is,

Rearranging gives the transmittance as


Substituting equation (1-39) gives

Spectral transmittance for path length S in a uniform gas volume

-7; (X,T,P,S)=-- - 1-a; (A, T, Pi S ) = exp (- akS) (1-47)

d3Qi, i

From equation (1-40) the intensities (as shown in fig. 1-10) can then be
related as

i: (S) = i: ( 0 ) r i (S) (1-48)

By analogous arguments, the directional total transmittance is given

by (it is again assumed the reflectance of the gas volume is negligible)
the following:

Directional total transmittance for path length S in a ungorm gas

volume = r ' (T, P, S) =1-a)'(T, P , S )

J i; (0) exp (- UAS) dX


EXAMPLE1-3: Some types of nuclear explosions produce, at their

peak, an emissive power spectrum like that of a blackbody at 6000 K.
The Sun also emits very close to this spectrum. Consequently, t h e
transmissivity of the atmosphere for solar radiation can be used to
determine the attenuation of energy from a nuclear explosion.
When the Sun is directly overhead, the total transmittance of the
atmosphere for solar radiation averages 35 percent throughout the fall
and winter in the Great Lakes region. Assume that a 20-megaton weapon

is detonated at a height of 10 km and dissipates its energy uniformly over

a period of 4 sec. Assume further that the fireball during this period is
1000 m in diameter and that 50 percent of the total energy is dissipated
as thermal radiation. Calculate the radiant energy flux directly below
the burst at ground level.
The total energy expended by the fireball per unit time is (1 megaton =

20 megatons = 5 X 1015 -
c a1
Q= 4sec sec

For 50 percent of the energy going into thermal radiation, the emissive
power of the fireball is

The intensity of radiation leaving the blackbody fireball is

and from equation (1-48) the intensity arriving at ground level is

To compute the energy arriving at the ground, the fireball is treated

approxiinately as a differential area with projected area dAp=~R$reb,ll
seen from the ground. Then the energy reaching the ground directly
below the fireball per unit time is

The energy flux at ground level directly below the fireball is

0.175Q -0.175 X 5 X 1015-- 69.5 cal

dQ' - --
4n;s2 471. 1012 (cm2)(sec)

Note that the result is independent of Rfirebarl.




cloud cover

1-12.-Areas receiving 15 cal/(cm2)(sec) or more of radiative flux as function of
~veathel.conditions during four large-yield high-altitude weapon detonations (ref. 12).

The rather grim result is that this flux applied over a few seconds is
more than four times that required to ignite newspapers. A more com-
plete approach to problems of this type is given in reference 12, where
the slant angle to the ground is included. Higher altitude detonations
than those of example 1-3 were studied. Figure 1-12 shows some results
of reference 12 not recommended for the imaginative reader.


It was tacitly assumed in volume I of this series that opaque solids
emit energy based solely on the temperature and physical properties of
the body. The spectrum of emitted energy was assumed unaffected by the
characteristics of any incident radiation. This is generally true because
all the absorbed part of the energy incident on an opaque solid is quickly
redistributed into internal energy states in an equilibrium distribution at

the temperature of the solid.

In a gas, the redistribution of absorbed energy occurs by various types
of collisions between the atoms, molecules, electrons, and ions that com-
prise the gas. Under most engineering conditions, this redistribution
occurs quite rapidly, and the energy states of the gas will be populated
in equilibrium distributions at any given locality. When this is true, the
Planck spectral distribution correctly describes the emission from a
blackbody, and equation (1-34) correctly describes the emission from
a gas volume element.
The assumption, that a gas will emit according to equation (1-34)
regardless of the spectral distribution of intensity passing through and
being absorbed by dV, is a consequence of the assumption of "local
thermodynamic equilibrium" or LTE. When the condition of LTE is
not present, the calculation of radiant transfer becomes much more
Cases where the LTE assumptio~lbreaks down are occasionally en-
countered. Examples are in very rarefied gases, where the rate and/or
effectiveness of interparticle collisions in redistributing absorbed radiant
energy is low; when rapid transients exist so that the populations of
energy states of the particles cannot adjust to new conditions during the
transient; where very sharp gradients occur so that local conditions de-
pend on particles that arrive from adjacent localities at widely different
conditions and may emit before reaching equilibrium; andwhereextremely
large radiative fluxes exist, so that absorption of energy and therefore
population of higher energy states occur so strongly that collisional
processes cannot repopulate the lower states to an equilibrium density.
Under any of these conditions, the spectral distribution of emitted
radiation is not given by equation (1-34). Then the populations must
be determined by detailed examination of the relation between the
collisional and radiation processes and their effect on the distribution
of energy among the various possible states- a most formidable under-
taking. It is, however, necessary in examination of shock phenomena
(sharp gradients), stellar atmospheres (extreme energy flux and low
density), nuclear explosions (transients, sharp gradients, and extreme
fluxes), and high altitude and interplanetary gas dynamics (very low
A gas with small optical thickness can have transmitted within it
radiation from regions at widely different conditions. For this reason, a
nearly transparent or "clear" gas is more likely to depart from LTE
than is an optically thick gas of the same density.
A very prominent no1rLTE effect is found in the laser. In this device,
a material with a met:stable energy state is excited by some external
means. Because the excited state is metastable and is also chosen so that

no competing process is trying to depopulate it, its population can reach a

value well above the equilibrium value. This condition is called apopula-
tion inversion. The material is then exposed to radiation containing
photons with the same frequency as the transition frequency from the
excited to a lower state in the material. This radiation induces or stimu-
lates the transition to the lower state. Consequently, a large number of
photons with the transition frequency are emitted, thus amplifying the
intensity of the incident radiation. This process leads to the acronym
light anlplification by the stimulated emission of radiation, or laser.
Such non-LTE problems are beyond the scope of this work. It will be
assumed here that LTE always exists and that although the flux arriving
at a volume element dV may come from localities at widely different tem-
peratures, the emission from dV will be governed by equation (1-34).


In this chapter, some basic concepts and definitions in the theory
and physics of gas radiation have been introduced. h succeeding chap-
ters, application of these concepts will be made to the problem of radiant
energy transfer in gases. The idea of induced emission has been intro-
duced, and the method of accounting for this effect in the equation of
transfer will be shown in chapter 2.

1. LANGLEY, S. P.: Experi~nentalDetermination of Wave-Lengths in the Invisible Pris-
matic Spectrum. Memoirs Nat. Acad. Sci., vol. 2, 1883, pp. 147-162.
2. GOODY, R. M.: Atmospheric Radiation. Vol. I-Theoretical Basis. Clarendon Press,
Oxford, 1964.
3. KONDRATYEV, K. YA.: Radiation in the Atmosphere. Academic Press, 1969.
4. GARDON, ROBERT:A Review of Radiant Heat Transfer in Glass. J. Am. Cer. Soc., vol.
44, no. 7, July 1961, pp. 305-312.
5. CHANDRASEKHAR, S.: Radiative Transfer. Dover Publ., 1960.
6. KOURGANOFF, VLADIMIR: Basic Methods in Transfer Problems; Radiative Equilibrium
and Neutron Diffusion. Dover Publ., 1963.
7. SIEGEL, ROBERT;AND HOWELL, JOHNR.: Thermal Radiation Heat Transfer. I-The
Blackbody, Electromagnetic Theory, and Material Properties. NASA SP-164, 1968.
8. PENNER, S. S.: Quantitative Molecular Spectroscopy and Gas Emissivities. Addison-
Wesley Publ. Co., 1959.
Theory of Gas Dynamics. Addison-Wesley Pub. Co., 1965.
10. BATES, DAVID R., ED.:Atomic and Molecular Processes. Academic Press, 1962.
11. HOWELL, JOHN R.; AND SIEGEI., ROBERT:Thermal Radiation Heat Transfer. II-Radi-
ation Exchange Between Surfaces and in Enclosures. NASA SP-164, 1969.
12. ATLAS, REYNOLD;AND CHARLES,B. N.: Atmospheric Attenuation of the Thermal
Radiation from a High-Altitude Nuclear Detonation. Paper 64-318, AIAA, June 1964.

13. KULANDER, JOHN L.: Nan-Equilibrium Radiation. Rep. R64SD41, General Electric
Co., June 1965. (Available from DDC as AD-61 7383. )
14. THOMAS,RICHARDN.: Some Aspects of Nonequilibrium Thermodynamics in the
Presence of a Radiation Field. University of Colorado Press, 1965.
15. GARBUNY, MAX: Optical Physics. Academic Press, 1965.
16. AROESTE,HENRY;A N D BENTON,WILI,IAMC.: Emissivity of Hydrogen Atoms at
HighTemperatures. J. Appl. Phys., vol. 27,1956, pp. 117-121.
17. MEYEROTT,R. E.: SOKOLOFF, J.: AND NICHOLLS,R. A.: Absorption Coefficients of
Air. Rep. LMSD288052. Lockheed Aircraft Carp. (AFCRC-TR-59-296). Sept. 1959.
18. HOTTEL,H. C.: Radiant-Heat Transmission. Heat Transmission. Third ed., William H.
McAdams, McGraw-Hill Book Co., Inc., 1954, ch. 4.
Chapter 2. The Equations of Transfer for an
Absorbing-Emitting Gas
In chapter 1 some of the basic concepts and definitions were presented
for intensity, emission, and absorption within a medium. The radiation
traveling along a path within a medium is attenuated by absorption and
scattering, and is enhanced by both spontaneous and induced emission
and also by radiation scattered in from other directions. The fundamental
processes of absorption and emission as discussed in chapter 1 will be
employed to develop a differential equation governing the radiation
intensity along a path through the absorbing and emitting medium.
This equation is called the equation of transfer. The effects of scattering
will be neglected in this chapter and in chapters 3 to 7; scattering will
be considered in chapter 8.
When obtaining a solution to the equation of transfer, a constant of
integration will be introduced; the evaluation of this constant introduces
the intensity at the origin of the radiation path being considered. Because
the origin is usually at the boundary of the radiating medium, the radia-
tion at the boundaries is thereby coupled into the radiation distribution
within the medium.
The intensity gives the radiation that is traveling in a single direction,
per unit solid angle, and that is crossing a unit area normal to the direc-
tion of travel. To obtain the net energy crossing an area, an integration
must be made that includes the contributions of the intensities crossing
in all directions. This results in an equation for radiative flux which will
be used in the formation of heat balances within the medium.

absorption coefficient
speed of light in a medium
spacing between parallel plates or diameter ratio
exponential integral function, eq. (2-45)
emissive power
photon distribution function
Planck's constant
radiation intensity

i, j, k unit vectors in x, y, z coordinate directions

n unit normal vector
P 'pressure
Q energy per unit time
9 energy flux, energy per unit area and time
r position vector
S coordinate along path of radiation
s unit vector in S direction
absolute temperature
radiant energy density
coordinates in Cartesian system
cone angle, angle from normal of area
circumferential angle
optical thickness
optical thickness for path of length D
Stefan-Boltzmann constant
solid angle
a sbsorbed
b blackbody
e emitted
i incident mean value, eq. (2-21)
P Planck mean value, eq. (2-20)
1, v spectrally dependent
f along directions having positive cos p
- along directions having negative cos /3
1,2 surface 1 or 2
directional quantity
+ true value, not modified by addition of induced emission
* dummy variable of integration
- averaged over all incident solid angles


The equation of transfer in a nonscattering medium will now be
derived. As stated in section 2.1, this will describe the intensity of

radiation at any position along its path through an absorbing-emitting

2.3.1 Derivation
Bouguer's law, equation (1-12), in a nonscattering medium accounts
only for attenuation by absorption. The equation of transfer is an exten-
sion of Bouguer's law to include the contribution to the radiation in-
tensity of energy emission along the path.
Consider radiation of intensity ii;(S) within a region of absorbing-
emitting medium as shown in figure 2-1. Attention will now be directed
to the change of intensity as the radiation passes through a distance dS.
Not including the gas emission for a moment, the intensity at S + dS
for a nonscattering gas is equal to the intensity at S plus the change
caused by absorption in dS; that is,

Using equation (1-8) with KA= ah ,in this case gives

Note that ah has been used in equation (2-1) rather than the "true"
absorption coefficient ac. Thus the intensity i ; ( S + d S ) is the result

emitting medium

2-1. -Geometry for derivation of equation of transfer.

387-309 0 71 -4

not only of true absorption, but also includes the contribution of induced
emission as discussed in section 1.5.5.
Assuming the radiation along the path is in local thermodynamic
equilibrium, the spontaneous emission contribution by the gas along the
path length dS to the intensity in the S direction is given by equation
(1-35) as

Adding equations (2-1) and (2-2) gives the intensity of the radiation
i{ (S+ d S ) as

i i ( S + d S ) = i L ( S ) cli;, .+ di;, ,
= i; ( S ) [ l - ah ( S )d S ] + ax ( S )i{, ( S )dS (2-3)

where i;(S d S ) now includes all emission from the gas as well as the
contribution of the attenuated incident intensity. The change in intensity
di; of the incident radiation as a result of passing through dS is then

A form of equation (2-4) in astrophysical texts and which is often

more convenient to work with is obtained by combining ahdS into a
single quantity, that is,

which takes into account the absorption coefficient and the differential
path through which the radiation is traveling. The d ~ is h called the
optical differential thickness. By integrating equation (2-5) as in equa-
tion (1-17) the optical thickness or optical depth is obtained for a layer
of thickness S or a path of length S

KA( S ) =/ S
ah ( S * )dS* (2-6)

Using the optical differential thickness results in equation (2-4) taking

the form

Equation (2-7) is the equation of transfer for an absorbing-emitting gas.

There is a basic advantage for including the induced emission in the
absorption coefficient. The induced emission as discussed in section
1.5.5 is in the same direction as the transmitted radiation. The spon-
taneous emission, however, is uniform over all directions. Thus by
combining the induced emission with the "true" absorption to form ax
(and KA),the quantities depending on the direction of the incident radia-
tion have been brought together. The resulting emission term in the
equation of transfer contains only spontaneous emission and hence does
not depend on direction.

2.3.2 Integration by Use of Integrating Factor

Equation (2-7) is a first-order linear differential equation and a general
solution can be obtained by use of an integrating factor. Multiplying
through by the factor exp ( K ~gives

- -- [ ~ ; ( K Aexp
) I
( K A ) = iib(Ki) exp (KA) (2- 8)

Integrating over an optical thickness from K A = O to K A ( S )gives

;;(.A) exp (KA)- i:(O) = lo K1

iib(K:) exp (K:)~K:

i; (KA)= i i (0) exp (- KA)-k

1: iLb(K:) eXp [- (KA
- K:)]~K:

where KT is a dummy variable of integration.

Equation (2-10) is interpreted physically as the intensity at o p t i c a l
depth K A being composed of two terms. The first is the a t t e n u a t e d
incident radiation arriving at K A (including, however, the contributioll
of induced emission along the path), and the second is the intensity
resulting from spontaneous emission in the S direction by all t h i c k n e s s
elements along the path and reduced by exponential attenuation b e t w e e n
the point of emission K: and the location KA. Equation (2-10) is t h e
integrated form of the equation of transfer. As derived, this e q u a t i o n

applies to the spectral intensity traveling in the positive K A direction.

Although equation (2-10) is a general solution to the equation of
transfer, the intensity cannot be obtained directly from it unless the
temperature distribution is known. The temperature will determine
the blackbody emission term iAb(KA) in the integral on the right side.
Also the temperature distribution is needed to determine the absorption
coefficient ah(S) so that the local optical depth K ~ ( S can
) be com-
puted from equation (2-6) and the physical coordinate S thereby related
to the optical coordinate K A . The temperature distribution depends on
conservation of energy within the medium which in turn depends on the
total absorbed radiation in each volume element along the path. This
total energy quantity will be obtained in the next section by utilizing
the intensity passing through a location and integrating over all incident
solid angles and all wavelengths. The energy equation and equation
(2-10) yield the necessary relations from which compatible temperature
and radiation intensity distributions can be found.

EXAMPLE 2-1: A black surface element dA is 10 cm from an element

of gas dV (fig. 2-2). The gas element is a part of a gas volume V that is
isothermal and at the same temperature T as dA. If the gas has an absorp-
tion coefficient a h of 0.1 cm-' at wavelength 1 p m , what is the spectral
intensity at A = 1 p m that arrives at dV along the path S from dA to dV?
Because element dA is black and at temperature T, the intensity
at S = 0 is i,( (0) = iib (T). Since the gas is isothermal, the emitted
blackbody intensity in the gas is i,& ( K A=) i 6 (T). Substituting into the
integrated equation of transfer (eq. (2-10)) gives

) iLb (T) exp ( - K A f

i,( ( K A= ) &
,i , (T) exp ( - K A )
exp (KT)dKT

FIGURE2-2. -Geometry for example 2-1.


After carrying out the integral, this. reduces to

The iLl,(T) is given by equation (2-lla) of volume I for a gas with

refractive index n = 1. The intensity arriving at dV along an isothermal
path from a black surface element at the same temperature as the gas
is thus equal to the blackbody intensity emitted by the wall and does
not depend on ax or S. The attenuation by the gas of the intensity emitted
by the wall was exactly compensated by emission from the gas along
the path from dA to dV.


Equation (2-10) is concerned with the energy at only a single wave-
length traveling in only a single direction within the medium. The tem-
perature distribution within the medium is governed by conservation of
energy which depends on the energy arriving at a volume element in
all wavelength regions and from all incident directions. Since equation
(2-10) depends on iib which is a function of local temperature, the in-
tensity equation is coupled to the energy conservation equation to obtain
the radiation intensity and temperature distributions.
To derive the energy conservation equation consider the energy
absorbed by a volume element dV within a medium as shown in figure
2-3. The energy absorbed from the incident intensity ii ( A , w , K A )that

Medium boundary

FIGURE2-3. -Geometry for derivation of energy conservation relation.


arrives within the incremental solid angle do is, by analogy with equation

The incident intensity ii (A, w, K A )is given by the equation of transfer

(eq. (2-10)) as

;;(A, w. K A ) = ~ ; ( Aw,
, 0) exp (-KA)

where i;(A, w, 0 ) is the spectral intensity directed toward dV from

the system boundary in the direction of dw.
The energy absorbed ("true" absorption reduced by induced emission)
by dV from all incident directions is found by integrating equation (2-11)
with respect to w

For convenience in writing the equations a little more compactly a mean

incident intensity i x , i(A) can be defined by

4 , (A) i;(A, w. K A ) C ~ (2-14)

Then equation (2-13) becomes

By integrating equation (2-15) over all wavelengths, the total energy

absorbed by dV from the radiation field is obtained as

The total energy emitted spontaneously from dV is obtained by using

ax in equation (1-34) and integrating over all wavelengths; the result is

This equation contains the assumption that clV is so small that all the
energy emitted by dV escapes before any can be reabsorbed within dV.

2.4.1 Radiative Equilibrium

For situations when all energy exchange mechanisms such as con-
duction and convection are negligible compared with radiation and no
transients in local temperature are occurring, the total emitted energy
from clV is equal to the total absorbed energy. This is termed radiative
equilibrium and is simply a statement of steady-state energy conserva-
tion in absence of any other exchange mechanism but radiation. Using
equations (2-16) and (2-17) radiative equilibrium gives

2.4.2 Some Mean Absorption Coefficients

As a result of the emission integral on the left of equation (2-18) it is
convenient to define the Planck mean absorption coeficient a p ( T , P ) as

The arl is the mean of the spectral coefficient when weighted by the
blackbody emission spectrum. It will prove useful when considering
emission from a volume, and in certain limiting cases of radiative transfer.
Substituting equation (2-19) into equation (2-18) results in energy
conservation in the form

Thus, if i*, i ( A ) is known at a position within the gas, equation (2-20) can

be solved for T at that location. The Planck mean (111 is convenient since
it depends only on the properties at dV. It can be tabulated readily and
is especially useful where the pressure is constant over the geometry
of the system.
As a result of the absorption integral on the right side of equation
(2-20), another type of mean absorption coefficient can be defined. This
is the incident mean (or modiJied Planck mean) absorption coefficient
a i ( T , P ) given by

1; o*(A, T , p ) i ~i .( h ) d A
ni ( T , P ) =
1; ZA, i(A)dA

However, such a definition for general use has little value. A tabulation
of ai would have to be carried out for all combinations of incident spectral
distributions and spectral variations of local absorption coefficients.
Except in certain very limited special cases, the work involved in such
a tabulation would not be warranted. Further discussion of the physical
interpretation of various mean absorption coefficients is given in section
In order to evaluate the influence of some of the many variables in
gas radiation problems, it is sometimes convenient to consider a simple
geometry. A plane layer is often used and there is a considerable litera-
ture for this geometry in both engineering and astrophysical publications.
The astrophysical interest (refs. 1 and 2) stems from the fact that the
atmosphere of the Earth and the outer radiating layers of the Sun can be
approximated as a plane layer.
The plane layer is illustrated in figure 2-4. The temperature and

FIGURE2-4.-Plane layer geometry.


properties of the gas vary only along the x coordinate. An arbitrary path
S within the gas is at angle p to the x direction. The optical depth ~ ( x )
is now defined along the x coordinate as

The relation between optical positions along the S and x directions is

given by

xlcos 13
r ' p~ ) =cos
K ( S ) = / ~ ' ~ ~ S * = / ~ a d (cos ~ /@ X ( pZ ) (2-23)
, ~ a d x * = cos

The equation of transfer (eq. (2-7)) is written along any path S and
thus the K in equation (2-7) are K(S).By use of equation (2-23), the trans-
fer equation is written in terms of K(X)as

COS P ---+ ~ ) ( [ K A ( xPI) = i;b [KA(X)]

A partial derivative is used to emphasize that ih, depends on KA(X)

and p. The equation of transfer in integrated form, equation (2-lo),

where all the K in equation (2-25) are ~ ( x )A. convenient substitution

that is often used is to let p = c o s 0. Then equation (2-25) becomes

~ ; ( K Ap ,) = i;(O) exp


A gas having an absorption coefficient that is independent of wave-
length is called a gray gas. From the discussion of gas property spectral

variations such as in connection with figure 1-2, it is evident that gases

are usually far from being gray. However, there are some instances when
gases may be considered gray over a portion of the spectrum. In other
cases such as when particles of soot or other material are present or
are injected into a gas to enhance its absorption or emission of radiation,
the absorption coefficient of the gas-particle mixture may act as if the
mixture were nearly a gray gas. In addition, examination of the radiative
behavior of a gray gas provides an understanding of many of the features
of a real gas without some of the complicating features that real gas
effects introduce. The gray gas is thus of some practical and theoretical
interest and has consequently received a great deal of attention in the
literature. The equation for local intensity and temperature will now be
written for a gray gas.

2.6.1 Transfer Equations

For a gray gas, K A is independent of wavelength and will be called K for
simplicity. Then the local total intensity in the gas can be found by
integrating equation (2-10) over all wavelengths to give

Using the definition of total intensity, which is

results in equation (2-27) becoming

For a gray gas ax is independent of A and equation (2-19) gives np= a h .

The condition for radiative equilibrium as given by equation (2-20)
reduces to the following equality at ally optical depth within the medium:

As was defined in equation (2-14), the relation between ii in equation

(2-29) and i r in equation (2-28) is an integration over all incident solid


so equation (2-29) becomes

Equations (2-28) and (2-31) give a set of relations coupling if ( K ) and

T ( K )that may be used to determine the temperature distribution within
the gas when boundary conditions are prescribed. The boundary condi-
tions are needed to supply if(0)in equation (2-28) for each path. In the
following section the relations given here will be applied to a plane layer
of gray gas between infinite parallel black plates.

2.6.2 Plane Layer Between Black Plates

Consider two black infinite parallel plates that are separated by a
gray gas with absorption coefficient a ( T , P). The gas is in radiative
equilibrium. As shown in figure 2-5, the lower plate is at temperature

FIGURE2-5.-Geometry for finding distribution of temperature in gray gas.


TI, and the upper plate is at Tz. The plates are separated by a distance
D. It is desired to obtain expressions for the temperature distribution
in the gas and the energy transfer between the plates.
Since the geometry is a plane layer, all the K in what follows are K ( X )
as defined in equation (2-22). The temperature distribution is found
from equation (2-31). The integration of i' over w is conveniently ex-
pressed in two parts for intensities i$ approaching d V from directions
with positive cos p (from plate 1 in this case, 0 < p < 90") and i'ap-
proaching from negative cos P directions (from plate 2 in this case,
90" < p < 180"). These intensities are shown in figure 2-5 and it is noted
that p is measured from the positive x direction. Then equation (2-31)
can be written as

where the notation

b dw denotes integration over all solid angles in the
hemisphere in directions from surface 1 and dw denotes integration
over the hemisphere in directions from surface 2.
The intensities in equation (2-32) are obtained from equation (2-25).
This gives for radiation traveling from wall 1 at angle P and in the
direction of positive cos /3

i : ( ~ ) = i : ( O ) exp - + 1"
i i ( ~ * exp
) [
0 s p I dK*
cos p

The i:(O) is the total intensity leaving wall 1 ; and since wall 1 is black,

The total intensity under the integral is related to the local gas tempera-
ture by

Substituting this into equation (2-33) gives

i:(~) =I { c r ~ exp
7~ f (x)cos p i
cos p-
T 4 (~
~ * exp
) [ -(K-K*)
cos p ] dK*}

where 0 =SP < 90". By a similar analysis, the intensity at angle P

(90°< p < 180") reaching optical depth K from the direction of plate 2
is (note that cos p is negative in this range)

'Ti- cos p

where x D = l a(x)dx.
The intensities in equations (2-34) and (2-35) are substituted into
equation (2-32) to yield the following integral equation for T4(~):

T ~ ( K )=. ir ,8 [ ~ cos p
T exp (-K )
'+ cos p I'
T 4 ( ~ * exp
) ] d ~ +*T$ exp [- (cosK pD ]

'+ cos p "J: T 4 ( ~ * exp

[ ( Kcos
* -L
3)]d~*}dp (2-36)

where the substitution dw = 2 7 ~sin B d p has also been made. Solutions

giving the temperature distribution will be discussed later.
The radiative energy flux in the positive x direction crossing the
plane at x in figure 2-5 is found in two parts, one from the i: and one
from the i'. Since intensity represents energy crossing an area normal
to the direction of i f , the projection of the area dA must be considered
normal to either iC or i'. The flux in the positive x direction from the
iL is

P+(K) =L=O rlz

i : ( ~ ) cos P 211 sin 0 d p (2-37a)

The flux in the negative x direction from the i' is

i ) ( ~ )cos (T-/3)2'~i- sin ('~i--P)d('~i--/3)

g - ( ~ =-27~
) IT
~ L ( Kcos
) p sin P d p (2-3713)

The net flux in the positive x direction is

9 ( ~=) 9+(K) -9-(K)


Substituting equations (2-37) into equation (2-38) gives

q(K) = 2 n [I: i : ( ~ ) cos p sin p rip+

IT;2 i i ( ~ cos
) p sin P
The intensities from equations (2-34) and (2-35) are substituted into
equation (2-39) and the integrals combined to yield

q ( ~=) 2 sin p cos P

I L;;)
crT4 exp -

+ cos p 0
T K )e x [- (K-K*)
cos p
d ~ * - crTi exp

- (K*- K )
/ y ~B ~ ( ~ * ) e x p [
- "cos cos p IdK*}@
For energy transfer only by radiation (radiative equilibrium) in the geom-
etry being considered here, q ( ~ must
) be independent of K because
there are no energy sources or sinks in the gas. Evaluating q at the con-
venient location K = O then gives the heat flux flowing from wall 1 to
wall 2,

T'(K*) exp cos p d ~ * d] p

+Lcos p 1 0
"" T'(K*) exp (s)
d ~ * dp
] (2-41)

This can be evaluated after T 4 ( ~is) found from equation (2-36).

For the limiting case as the absorption of the medium between the
plates becomes very small KI) -+ 0, equation (2-41) reduces to

which is the correct solution for black infinite parallel plates separated
by a transparent medium. In addition for this limit equation (2-36) yields

so that a nearly transparent gray medium approaches a temperature to

the fourth power equal to the average of fourth powers of the boundary
Solutions for the temperature distribution in a gray gas with tempera-
ture-independent properties contained between infinite parallel plates
have been obtained by many researchers. Some of the solution methods
will be discussed in succeeding chapters. Heaslet and Warming (ref. 3)
have presented solutions accurate to four significant figures for the
quantities [T4( K )- Ti]/ (T: - T:) and q/ [ o ( T : - T:) 1. Their results
for the black boundary case are shown in figure 2-6.
The temperature distributions of figure 2-6(a) show that a discon-
tinuity exists between the wall temperature and the gas temperature at
the wall. This phenomenon is called the temperature "slip" or "jump."
If the slip were not present, the curves would all go to 1 at K / K D = O
and 0 at K/KD = 1. The slip disappears when heat conduction is included
in the analysis. To determine the magnitude of the slip, the gas temper-
ature is evaluated at K = 0. This gives by using equation (2-36)

)' IT''
T ~ ( K = o =-
2 0
sin [T:+T; exp (--)- KI)
cos p

-k -
cos P
T ~ ( *)
K exp (=)- K * *] d p

which can be written as

Again it is seen that as Kg approaches zero, equation (2-42) reduces


Optical thickness,

Relative optical depth, K/KD

Optical thickness, KD

(a) Temperature distribution.

(b) Energy flux.
2-6.-Temperature distribution and energy flux in gray gas contained between
infinite black parallel plates (ref. 3).

0 .5 1.0 1.5 2.0 2.5 3.0
Optical thickness, KD

2-7.-Disconti~~uity at wall between gray gas and black wall temperatures (ref. 3).

The magnitude of the slip for a gray gas with constant absorption coef-
ficient is shown in figure 2-7 as a function of the optical thickness of the
layer. (Note that from symmetry T: - T " ( K =0) = T"K= K D ) - T i . )

2.6.3 Use of Exponential Integral Functions

There are some mathematical substitutions that are useful in deal-

ing with the results derived from the equation of transfer for plane layers.
By letting p= cos p, equations (2-36) and (2-41) become

T*( K ) =; 1'{ (?)Tj exp

T * ( K * )exp d~*+
T: exp [ -(KD-K)
+llKKD [-'"I-"']
T * ( K*) exp dK*}dp

The exponential integral function can now be introduced. This func-

tion is defined as

E,,([)= 1 pn-2exp (T) dp

'Then equation (2-43) can be written as

and equation (2-44) becomes

The exponential integral functions are discussed in detail by Kourganoff

(ref. 1) and Chandrasekhar (ref. 2). For convenient use by the reader
some of the important relations are given in the appendix.


The radiation field and transfer of radiation in a medium can also
be expressed in terms of a photon model. This is sometimes helpful
in providing a physical picture of the transport and is also useful in
Monte Carlo methods as will be considered in chapter 6. Since photon
energy is related to the frequency of the radiation, frequency will be
used in this section rather than wavelength.
When considering the radiation as a collection of photons the condi-
tions at any location in the medium are given by the photon distribution
function$ Let
f (v,7, S)dvdVdw (2-48)

be the number of photons traveling in direction S in frequency interval

dv centered about v, in volume dV at position 7,and within solid angle
dw about the direction S (see fig. 2-8(a)). Each photon has energy hv.
The energy per unit volume per unit frequency interval is then hvfdw
integrated over all solid angles. This is called the spectral radiant energy

To obtain the intensity, the energy flux in the S direction is needed

across the area dA in figure 2-8(a) which is normal to the S direction.
The photons have velocity c, and the particle density traveling in the
normal direction across dA is fdvdo. The number of particles crossing
dA per unit time is then cfdvdwdA. The energy carried by these particles
is I~vcfrlvdwclA.The spectral intensity is the energy per unit time, unit
frequency interval, and unit solid angle crossing a unit area normal to
the direction of the intensity. This gives the intensity at location ;and
in direction S as
i:= hvcf ( v , 7, S ) (2-50)

(a) Quantities in intensity derivation. (b) Quantities in flux derivation.

(c) Spherical coordinate system for radiative flux vector.
2-8. -Geometries used in derivations of radiative energy quantities.

The energy density; and the intensity can then be related by using
equation (2-50) to eliminate f from equation (2-49), that is,

This integral was encountered in equation (2-31).

Now consider the energy flux crossing an area within a medium. As
shown in figure 2-8(b) let d A be an arbitrary element whose unit normal
vector is ii,. Energy passes through d A from all directions; a typical direc-
tion is the S direction at angle p to ii. This energy is given by hvcfdvdwdA
cos p. To obtain the net energy flux crossing d A integrate over all in-
cident solid angles. The energy per unit of dA moving across dA in the
direction of increasing 6 is then (note that cos /3 becomes negative for
p > n/2 so that the sign of the portion of the energy flux traveling in the
direction opposite to the positive Ji direction is automatically included)

f cos /3 dw = dv i: cos /3 dw (2-52)

The latter form of equation (2-52) was obtained by use of equation (2-50).
Let S be a unit vector in the directions of the photons. Then cos P=s'.ri
and equation (2-52) can be written as

Thus dq, is the component in the 7z direction of a flux vector given b y

that is, dq,= ii . dq',.

To further reveal the vector nature of dij,,, consider a spherical
coordinate system as shown in figure 2-8(c). The unit vector 3 can then
be written as

s'=? cos 0 sin P + j sin 0 sin p + cos /? (2-55)

Substituting s and dw = sin P dpd0 into equation (2-54) gives the vector
dij, in terms of its three components,

i: ( p , 0 ) cos I9 sin2P dodo

+i 1'-
s=o p=o
il ( P , 19) sin I9 s i n 2 PrlprlI9


The equation of transfer has been derived which gives the variation
of intensity for radiation traveling in a fixed direction through an
absorbing-emitting medium. The equation expresses how the intensity
is attenuated by absorption and strengthened by emission; scattering
has been neglected. The equation of transfer was integrated so that the
intensity can be found along a path if an initial intensity is known; the
initial intensity would usually be that leaving a boundary. Since the
intensity involves the emission along a path, and the emission depends on
temperature, the temperature distribution in the gas is required to eval-
uate the intensity solution.
The gas temperature distribution is found from the energy conser-
vation equation. The energy terms are obtained from the total radia-
tion fluxes which are found by integrating the energy carried by spectral
intensities over wavelength and direction. Thus the solution of the equa-
tion of transfer is coupled to the energy equation and it was found that
the temperature distribution is governed by an integral equation.
Chapter 3 will consider in detail various approximate solutions that
have been obtained by use of the equation of transfer.

1. KOURGANOFF,VLADIMIR: Basic Methods in Transfer Problems. Dover Publications,
Inc., 1963.
2. CHANDRASEKHAR, SUBRAHMANYAN: Radiative Transfer. Dover Publications, Inc., 1960.
3. HEASLET, MAXA.; AND WARMING,ROBERTF.: Radiative Transport and Wall Tempera-
ture Slip in an Absorbing Planar Medium. Int. J. Heat Mass Transfer, vol. 8, July
1965, pp. 979-994.
Chapter 3, Approximate Solutions of the Equation of
Exact solutions to the radiative transfer equations to yield tempera-
ture distributions and heat flows in an absorbing-emitting medium require
considerable effort in most practical cases. Two approaches can be
taken to circumvent this complexity. In the first, the equation of transfer
may be simplified either by neglecting one or more terms when justified
or by transforming it into a diffusion equation. In the second, the com-
plete equation of transfer is used but approximate solutions are obtained
to it.
The equation of transfer was derived and integrated in chapter 2 to
give the variation of intensity along a direction of propagation in an
absorbing-emitting medium. The assumptions used in chapter 2, that
will also be retained here, are that there is no scattering, no heat con-
duction or convection, and that the gas is in local thermodynamic
equilibrium. The expression for the spectral intensity was given by
equation (2-10). This showed that the intensity along a path depends
on the intensity at the origin of the path, for example at a boundary, and
on the temperature distribution along the path.
There are three approximate solution methods that involve neglecting
terms in the transfer equation: the transparent, emission, and cold
medium approximations. They are summarized in table 3-1 and will be
treated in section 3.3 of this chapter.
The last approximation in table 3-1 is the diffusion approximation.
This is not obtained by neglecting a term in the equation of transfer,
but rather is derived by transforming the integral equations for the
radiative energy balance into a diffusion equation. The details of the
derivation, the approximations that are involved, and the solution of
the resulting diffusion equation form the second main portion of the
chapter, section 3.4.
The remainder of the chapter will deal with approximate solutions
of the complete equation of transfer. These are methods such as the
Milne-Eddington approximation and the differential approximation.

I Approximation I Form of equation of transfer

I Conditions
Strong i i ( S )= i l ( 0 ) The medium has such a low
transparent absorption coefficient that
an intensity does not
change by absorption or
emission while traveling
within the medium.

Emission No energy is incident from

the boundaries, and the gas
is relatively transparent
I so that emitted'energy

I from the gas passes

within the system without
significant attenuation.
Cold medium i:(S) = i i ( 0 ) erp [- o * ( ~ * ) d ~ * Emitted
] radiation from
medium is negligible
compared to that incident
from boundaries or
external sources.

Diffusion The optical depth of the

gas is sufficiently large,
and the temperature
gradients sufficiently small
so that the local intensity
results only from local

coefficients in eq. (3-105)
absorption coeficient
constants in Planck's spectral energy distribution
spacing between parallel planes, or diameter
ratio, (1- E ) / E
emissive power
volumetric energy generation rate
length over which temperature changes significantly
mean absorption value defined b y eq. (3-50)
radiation intensity

direction cosines
absorption mean free path, llax
spherical harmonics, eq. (3-107)
energy per unit time
energy flux, energy per unit area and time
sphere radius
radial coordinate
position vector
coordinate along path of radiation
unit vector in S direction
absolute temperature
distances measured along Cartesian coordinates
functions of angle in eq. (3-105)
cone angle, angle from normal of area
Gamma function
Kronecker delta
hemispherical emissivity
circumferential angle in fig. 3-3(b)
optical depth
optical thickness for path length D
Stefan-Boltzmann constant
temperature ratio, tables 3-11 and 3-111
dimensionless heat flux, tables 3-11 and 3-111
function defined by eq. (3-34)
solid angle

mean absorption coefficient in eq. (3-49)
evaluated at interface between gas regions 1 and 2
Planck mean value
Rosseland mean value in eq. (3-39)
net value in r direction
evaluated on the wall
net value in the z direction
propagating in positive or negative z direction, respectively

A spectrally dependent
Ah value integrated over wavelength interval Ah
0 evaluated at point of origin, initial value
1, 2 boundary 1 or 2, respectively, or region 1 or 2, respectively
+, - propagating in positive or negative direction
I directional quantity
* ** 7 dummy variable of integration
- average over all incident solid angles
( 0( 12 zeroth-, first-, or second-order term or moment


In chapter 2 the equation of transfer was integrated to give the in-
tensity variation along the optical path K (eq. (2-10)). For use in the
present discussion this equation is repeated here in terms of the actual
distance S along the path

i;(S) =i;(O) [-los

exp ai(~*)&*]

+los ~ A *)i;,
( S (S *) exp [-I:' aA(s**)&**] ds* (3- 1)

The intensity i;(S) as given by equation (3-1) depends on the intensity

i; (0) leaving the boundary at S=O, and on the temperature distribu-
tion since the local temperature governs the variation of ii, and a A .
As given by equation (2-20), the energy conservation equation needed to
obtain the temperature distribution involves an integration of incident
intensities from all solid angles. This results in a coupling of the energy
conservation and transfer equations which can become quite complex.
It is often possible to use some approximations that provide consider-
able simplification. Three of these approximations as summarized by
the first three entries in table 3-1 will now be discussed.

3.3.1 The Transparent Gas Approximation

When the optical depth along a path in the gas is small, the inte-
grated form of the equation of transfer can be simplified as the two
exponential attenuation terms in equation (3-1) each approach unity.
The intensity equation (3-1) then reduces to

There is no attenuation along the path of either the energy emitted in

the gas or the energy that enters at S=O. In sonle instances, an even
stronger assumption can be made. If the absorption coefficient is suf-
ficiently small and the i i ( 0 ) is finite, the emission from the gas as
given by the integral in equation (3-2) becomes negligible compared
with the intensity at S = 0 and equation (3-2) reduces simply to

i; (S) = i; (0) (3-3)

This is the strong transparent approxin~ation listed in table 3-1. An

incident intensity is thus essentially unchanged as it travels through
the gas. The local energy balances based on this simple intensity re-
lation are obviously much easier to carry out than those involving the
complete equation of transfer. The use of the strong transparent approxi-
mation will now be demonstrated.

EXAMPLE3-1: Two infinite parallel black plates at temperatures TI

and T S as in figure 2-5 are separated by a small distance D, and the space
between them filled with a gas of absorption coefficient ah. Assuming that
the strong transparent approximation holds, derive an expression for
the gas temperature as a function of position between the plates. It is
assumed the gas is in local thermodynamic equilibrium although this
assumption can sometimes break down in a thin gas a s mentioned in
section 1.8.
Equation (2-20) is the general expression for local radiative equilib-
rium in the gas. For the present case, it gives

As in equation (2-32) and from the definition (eq. (2-30)), t i , i is given

by the contributions approaching a volume element from above and
below, that is,

4nih, i ( ~ ) i; (A, o ) d o = j Q i;, ( A ,o)do+/
i;- (A, o ) d o (3-5)

Since the walls are black, the strong transparent approximation gives


Then since the black intensity is independent of angle, equation (3-5)


47rih, i ( A ) =2niL6(A, T I ) / 7 r ' 2 sin p dp+ 2ri;,,(A, T 2 ) sin P c@

Substituting equation (3-6) into equation (3-4) gives at any x position

between the plates

(rT4( x )=- " , I )+ iL6(h,T i ) l d A

l r n a ~ (xh) ,[ i k b ( A T
~ u / J ( x 0)

Equation (3-7) can be solved iteratively for T ( x ) . The necessity for an

iterative solution arises because ah depends on local temperature.
If a i does not depend on gas temperature, then using equation (2-19)
results in


Then equation (3-7) reduces to

T4( x )= [a~~(T~)T~+o,~(T~)T~]
2 a ~ [ T ( Ix )

The local temperature solution, although still requiring an iterative solu-

tion on T ( x ) and a p [ T ( x ) ] is
, relatively easily found by use of tabulated
values of ap(T). Note funher that for a gray gas with temperature-
independent properties, up is a constant and equation (3-8) reduces still
further to

Hence, the entire gas approaches a fourth-power temperature that is the

average of the fourth powers of the boundary temperatures. This limit
was also found in section 2.6.2 following equation (2-41).

3.3.2 The Emission Approximation

In the strong transparent approximation, the gas was optically thin and
the local intensity was dominated by the intensity incident at the gas
boundary. In the emission approximation tlie gas is again optically thin
but there is also negligible incoming energy at the gas boundary from
external sources. For these conditions, in equation (3-1) the exponential
attenuation terms both become unity as ah is small, and iL(0) is zero.
This gives
i;(S) =
P ax(S*)iib(S*)dSX (3-10)

Thus the only important term is the emission which is of order axi,&.
Reabsorption of emitted energy along a path is of order ax(c~xi{~,)
and is
negligible because ax is very small. The intensity i ; ( S ) in equation
(3-10) is consequently the integrated contribution of all the emission
along a path a s the emitted energy travels through the gas without
Equation (3-10) can be integrated over all wavelengths to give tlie
total intensity

i t ( S )= 1"i;(~)o'h=% [l' o h ( ~ * ) i ; b ( ~ * ds*


. (2-19)) is now applied to give

The definition of a ~ (eq.

The fact that equation (3-11) contains the Planck mean absorption
coefficient and was derived for optically thin conditions has sometimes
led to the statement that the Planck mean absorption coefficient is appli-
cable only in optically thin situations. However, the Planck mean was
defined in general for the emission from a volume element in connection
with equation (2-18), and hence can be applied for the emission term in
a gas of any optical thickness.

(a) Slab geometry for example 3-2. (b) Emission from spherical gas-
filled satellite with
transparent skin.
3-1. -Examples for emission approximation.

EXAMPLE3-2: Find the flux emerging from an isothermal slab of gas

with Planck mean absorption coefficient 0.010 cm-' and thickness
D = l cm if the slab is bounded by transparent nonradiating walls
(fig. 3-l(a)).
If i t @ ) is the emerging total intensity in direction p , the emerging
flux is

Since the slab is isothermal with constant up, equation (3-11) can be
integrated over any path Dlcos p through the slab to yield

aT4 D
i l ( P ) = a p --
n. cos p
apc+T4Dsin p dp= 2apaT4D

Substituting the numerical values gives

It should be realized that equation (3-12) is really not a precise result

even though the slab is optically thin in the sense that the optical thick-
ness based on D is apD=O.Ol % 1. This is because some of the radia-

tion reaching the slab boundary has passed through the thickness
Dlcos p. Because for large p directions this path length becomes in$nite,
the emission approximation cannot hold. A more accurate solution of
the equation of transfer including the proper path lengths gives

which is 10 percent less than equation (3-12).

EXAMPLE3-3: An inflated spherical balloon satellite of radius R is

in orbit in the Earth's shadow. The satellite has a perfectly transparent
wall and is filled with a gray gas of constant absorption coefficient a , such
that aR 4 1. Neglecting radiant exchange with the Earth, derive a rela-
tion for the initial rate of energy loss from the satellite if the initial
temperature of the gas in the balloon is TO.
From the emission approximation equation (3-11), figure 3-l(b) shows
that the following can be written for the intensity at the surface:

since a and To are constants. From the geometry

S=2R cos fi

Then q, the flux leaving the surface, is

=I IOnpr
cos2 p sin p
~ T ~ R

To obtain the energy loss Q per unit time from the entire sphere,
multiply through by the surface area of the sphere

where V , is the volume of the sphere. This is what is expected-it was

found that any isothermal gas volume radiates according to this formula
(see section 1.6) so long as there is no internal absorption; the emission
approximation gives a compatible solution.

3.3.3 The Cold Medium Approximation

The final approximate form of the equation of transfer to be discussed
in this section is found when the local blackbody emission within the
medium is very small. Such a situation might arise in considering radia-
tive transfer within a cold medium such as an absorbing cryogenic fluid.
The integrated equation of transfer (3-1) reduces to

i; ( S )= i;(O) exp [- 1: a, (s*)ds* ]

The local intensity thus consists only of the attenuated incident intensity.

EXAMPLE3-4: 100 watts of radiant energy leave a spherical light bulb

enclosed in a fixture having a flat glass plate as shown in figure 3-2. If
the glass is 2 cm thick and has a gray absorption coefficient of 0.05 cm-I,
find the intensity leaving the fixture at an angle of 60" to the bulb axis.
(Assume the bulb diameter is 10 cm.)
An integration of equation (3-15) over A and S results in the total
it ( P ) = it ( 0 ) exp (- U S )

To obtain il(0),consider the light bulb a diffuse sphere. The energy

flux (emissive power) at the surface of the sphere is 100 watts (W)
divided by the sphere area. The intensity is this diffuse emissive power
divided by T.Then

3-2.-Intensity of beam from light fixture (example 3-4').

i t ( P ) = 0.101 exp (- 0.05 X

Note that this problem has involved only a simple attenuated trans-
mission solution. The cold-material approximation is just that, a neglect
of emission along the path' of transmission. In this problem it was
assumed that emission from the glass pIate is small, so that only the
intensity originating at the source needed to be considered.


When a medium is optically dense, the radiation within it can travel
only a short distance before being absorbed. Consider the situation where
this radiation penetration distance is small compared with the distance
over which significant temperature changes occur. Then a local intensity
will be the result of radiation coming only from nearby locations where
the temperature is close to that of the location under consideration.
Radiation emitted by locations where the temperature is appreciably
different will be greatly attenuated before reaching the location being
For these conditions it will be shown that it is possible to transform
the integral-type equations that result from the radiative energy balance
into a diffusion equation. The diffusion equation is like a heat conduction
equation. The energy transfer depends only on the conditions in the
immediate vicinity of the position being considered and can be described
in terms of the gradient of the conditions at that position. The use of the
diffusion approximation leads to a very great simplification in treating
many problems of radiative transfer. Standard techniques, including
well-developed finite difference schemes, can be used for solving the
resulting diffusion differential equations. Such methods for differential
equations are developed to a much higher degree and are more familiar
to most engineers, for example in the solution of heat conduction prob-
lems, than are the methods of solution for the corresponding integral
As will be shown in the derivations that follow, the diffusion approxi-
mation requires that the intensity within the medium be nearly isotropic.
This can occur well within an optically thick medium with small tempera-
ture gradients but cannot be valid near certain types of boundaries. For
example, at a boundary adjacent to a vacuum at absolute zero tempera-
ture radiation will leave the medium but there will be none incident
from the vacuum. As a result of this large nonisotropy, the diffusion
approximation will not be valid near this type of boundary. The ability

to apply diffusion methods at the interfaces between regions has been

improved recently by the introduction of so-called "radiation slip" or
"jump" boundary conditions.
For real gases there are wavelength regions that are essentially trans-
parent. The diffusion approach can only be applied at specific wave-
lengths or in wavelength bands for which the optical thickness of the
medium is greater than about two; the fact that some mean optical thick-
ness meets this criterion is not sufficient. Wavelength band-type applica-
tions of the diffusion method can be made in the optically thick regions.

3.4.1 Simplified Derivation of the Diffusion Equation

First, a simplified derivation of radiation diffusion in a one-dimensional
layer will be carried out to give the spirit of the diffusion approximation.
The diffusion assumption is made that the medium in question has an
absorption coefficient sufficiently large so that the absorption mean free
penetration distance l l n h is small compared with the distances over
which significant temperature changes occur.
Consider the layer of gas shown in figure 3-3(a). The equation of
transfer from equation (2-7) is

dsJ I Medium with

i absorption
coefficient ah(x)

(a) One-dimensional plane gas layer. (b) General three-dimensional region.

3-3. -Geon~etry for derivation of diffusion equations.

By using the relation dS= dxlcos P, the equation of transfer giving the
change of iL with x for a fixed P is written as

where the blackbody intensity does not depend on angle. Let H be a

length over which the temperature changes by a significant amount.
Then nondimensionalize equation (3-17) using H and let l / a i = la1 from
equation (1-16)

-- cos fi M = , - iLb(wx)
i ~ ( yfi)


We now proceed to obtain a solution to equation (3-18) in the form of a

series. The diffusion approximation states that the penetration distance
is small compared to the length over which the temperature changes
appreciably. Hence L,,,/H 6 1 and the intensity can be written a s a series
of functions iL(I1)multiplied by powers of L,,,/H

Substituting into equation (3-18) gives (up to terms in first power of l,,,/H)

Equating the zeroth-order terms gives

Equating the terms in l,,,/H and then using equation (3-21) to eliminate
i;(") gives
ii(l)=- cos p- diLb (3-22)

Equations (3-21) and (3-22) are substituted into the series solution
equation (3-19) to yield (up to terms in /,,,/H)

cos /3 di$
a h dx

This result reveals the important feature that i n the d i f u s i o n solution

the local i n t e m i t y depends o n the magn.itude and gradient of the
local blackbody intensity.
The local spectral energy flux at x flowing in the x direction is found
by multiplying ii by cos P dA and integrating over all solid angles as in
equation (2-39)

i; ( x , 0 ) cos /?sin P dp

= 27r 1
cos8= - 1
i ; ( x , cos p) cos /3 d(cos p ) (3-24)

Using equation (3-23) in equation (3-24) gives, after noting that i i b

does not depend on P,

Equation (3-25) is an important result known as the Rosseland diffusion

equation. This equation relates the local energy flux to local conditions
only; it does not involve integrals of contributions from other regions
and thus provides a considerable simplification over the exact formulation
of the equation of transfer.

3.4.2 The General Radiation Diffusion Equation

In the previous section the diffusion equation was derived for a simpli-

fied case. Only first-order terms were retained in the series of equation
(3-19) and an unbounded region of gas was considered. The genera1
equations used in radiation diffusion will now be derived including the
second-order terms. Boundary conditions will be introduced into the
theory so that the diffusion equations can be applied to finite regions. As
shown by the solution in section, the boundary conditions must
account for a jump in emissive power between the wall and the gas at the
wall for a situation where there is radiative transfer only. The derivation
follows the general outline of that of Deissler (ref. 1). The intermediate
equations in the following derivation become somewhat complex be-
cause of their general form. The final equations however are relatively
simple and are very useful. The Rosseland equation for local radiativefiux. - Consider the
geometry in figure 3-3(b). There is a volume element dV0 at xo, yo, zo
having cross-sectional area dAo in the x , y plane. The energy flux cross-
ing dAo originates from all surrounding volume elements such as dV. If
the emission from dV produces an intensity i i ( S ) , then the intensity
reaching dVo is given by equation (1-21) as

i i ( 0 ) = i i ( S ) exp [ - a ~ ( h ) S ] (3-26)

This accounts for attenuation along S but does not include emission
along S which will be accounted for later by integrating the contributions
of equation (3-26) from all elements of the volume. Note that a spatially
constant ax has been used. This is not restrictive here as in the diffusion
approximation the gas temperature does not change significantly over the
region contributing significant radiation to a location. The solid angle
subtended by dV when viewed from dAo is dA/S2 where dA is the pro-
jected area of dV normal to S. The energy per unit time incident on dAo
as a result of the intensity in equation (3-26) is then

d'Qh,i(O) = i i ( S ) exp [- a ~ ( h ) S-s;-
] dAo cos P dh (3-27)

From equation (1-35), the spectral energy emitted spontaneously by

dV per unit time is

Substituting equation (3-28) into equation (3-27) gives

i ( ,O ) = a ~ ( h ) i ; ~ T)dS
~'QA ( h , exp [ - - a ~ ( h ) S ] dAo cos /3 dA (3-29)

If as a result of the gas being optically dense, the radiation field at

dVo originates only from locations close to clVo, then iib ( A , T ) in equation
(3-29) can be expanded in a three-dimensional Taylor series about the
origin S=O in the hope that truncation of the series after a few terms
will give an adequate representation of the iLb distribution near dVo.
The general Taylor series in three dimensions can be written as

This general series will be carried for the next several steps and will then
be truncated to a few terms. By applying the binomial theorem twice to
expand the factor in square brackets, equation (3-30) becomes

This relation is substituted into equation (3-29), which is then integrated

over the half space encompassing positive z values. This gives all energy
traveling in the negative z direction that is incident on dAo as

X (S sin 0 sin p ) p - S (S sin /3 cos O ) S cos P sin p

X exp [-- ak(A)S] dSdpd0 (3-32)

where the integral and summation signs have been interchanged and
spherical coordinates of the following form have been introduced:

x-xo=S sin p cos 0

y-yo=SsinPsin 0
2-zo=S COS p
The solid angle dA/S%as also been replaced by sin /3 d p dB. Note that the

following assumptions have been used in the integration over the entire
half space in equation (3-32): (1) the ax is constant within the region
that contributes significantly to the energy flux at dAo,and (2) there are
no bounding surfaces that contribute significant radiation energy at
d o . Otherwise, the ax would have to be retained as a variable in the
integration, and the integration would have to be over a finite region with
a specified intensity along the boundaries.
Carrying out the integration of equation (3-32) gives


and r is the gamma function.
A similar derivation for the energy incident on dAo from below, that
is, energy traveling in the positive z direction, gives

The net energy flux passing through dAo in the positive z direction
is then

Similar relations can be derived for the x and y directions.

In the diffusion approximation a region is being considered where
temperature changes slowly with optical depth. Hence derivatives such

as ( l / a ! ) ( a l i i ~ , / a ~become
~l) small as n is increased and the series in
equation (3--36) can be truncated. Retaining only terms th,rouglz the
second derivative causes the formidable looking equation (3-36) to
reduce to

This is the general relation for local energy flux in terms of the emis-
sive power gradient and is in agreement with equation (3-25); it is the
Rosseland difusion equation for radiative energy transfer. Retaining
only first-order derivatives, as in the derivation of equation (3-25),
also gives this same equation because the second-order terms have been
found to cancel. Note that equation (3-37) has the same form as the
Fourier law of heat conduction. This allows solution of some radiation
problems by analogy with heat conduction methods.
To obtain the energy flux in a wavelength range, integrate equation
(3-37) over the wavelength band Ah (the parentheses and 0 subscript
will be dropped for simplicity)

This defines the mean absorption coefficient aR,AA as

By multiplying the numerator and denominator by az/aeb, this can be

written as

The aK is called the Rosseland mean absorption coeficient after S .

Rosseland who first made use of the diffusion theory in studying radiation
effects in astrophysics (ref. 2). The aeXb/aebcan be found by differentiating

Planck's law (eq. (2-llb), vol. I) after letting T = (eb/a)'I4 The emissive power jump as a boundary condition.. -Up to

now the position considered in the gas was sufficiently far from any
boundary so that the effect of the boundary did not enter the diffusion
relations. Now the interaction of the radiating gas with a diffuse wall
will be considered. Let the wall bounding the gas from above as shown
in figure 3-4 have a hemispherical spectral emissivity EAluz. All quantities
pertaining to the wall itself will have the subscript w to differentiate
them from quantities in the gas at the wall which will have now subscript.
Consider an area d A 2 in the gas parallel to the wall and immediately
adjacent to the wall. The spectral energy passing through dAz in the
negative z direction is

where the terms on the right account for the emitted and reflected
energy from wall 2, respectively. The net spectral flux across dA2 in
the positive z direction is then

This can be placed in the form

Now equation (3-35) is substituted for (d2Qh,+,)z. The first term of

equation (3-35) for n = 0 and dAo = d A z is

3-4. -Geometry for derivation of energy jump condition at opaque boundary.

Then equation (3-43) becomes at dAz in the gas adjacent to the wall

Retaining only terms through second order and using equation (3-37) to
remove the first derivatives in terms of the spectral energy flux result
in the following relation for the jump in emissive power at dAz

All the quantities that do not have a w subscript are evaluated at dA2
which is i n the gas adjacent to the wall. The quantities with a w subscript
are evaluated on wall 2 and d q ~ ,is the net flux in the positive z direction.
In a similar fashion, the jump in emissive power at dAl in figure 3-4 is

where the quantities with a w subscript are on wall 1, and those without
a w are i n the gas adjacent to wall 1.

Equations ( 3 4 5 ) and (3-46) are boundary conditions that relate the

emissive power in the gas immediately adjacent to the wall eAb to the
wall emissive power ehb,. It is evident that there is a jump in emissive
power in passing from the gas to the wall at each boundary. Some
applications to clarify the use of these relations will be given in sec-
tion 3.4.3. The use of equation (3-35) in the derivation of these boundary
relations assumes that the proportionality between local radiative fEux
and emissive power gradient i n the gas holds even at points i n the gas
very near to a bounding surface. Although this is not strictly true, the
use of the jump boundary conditions corrects to a good approximation
for the wall effects.

To apply the diffusion equation (3-38) in a wavelength interval, the

jump boundary conditions (eqs. (3-45) and (3-46)) must also be inte-
grated over an increment of wavelength Ah. The wall emissivities are
assigned average values in this range, and the integration is carried out
as in reference 1 to yield

where qAA
= lAX dq~.

In these equations, two mean coefficients are given as originally

derived in reference 1

The IAA has units of length to the fourth power times inverse energy
rate. The emissive power jump between two absorbing-emitting
regions.-When internal sources or sinks for energy are present in
absorbing-emitting media, it is possible in the absence of energy conduc-
tion to have a discontinuity in emissive power at the interface of two
such adjacent media. This is obtained by considering a volume element at
the interface between two regions. The lower region has absorption
coefficient ail and the upper axz. Then the net flux passing through the
element per unit area nornlal to z is by use of equations (3-33) and (3-35)
in media 2 and 1, respectively,

Neglecting terms of order higher than two gives the emissive power
jump as

The integrated form of equation (3-52) for a wavelength interval is given

in reference 3.
As will be shown in section, the value of the jump ex@.- eAbl from
this equation will be nonzero under certain conditions. Summary. -The general radiation diffusion equation, given at
a single wavelength by equation (3-36) or (3-37) and for a wavelength
band by equation (3-38), has now been derived. The general boundary
conditions at solid boundaries with normals into the gas in the negative
and positive coordinate directions are given at a single wavelength by
equations (3-45) and (3-46), respectively, and for a wavelength band by
equations (3-47) and (3-48). Finally, a boundary condition for use at
the interface between two absorbing-emitting media in the absence of
heat conduction is given by equation (3-52).

3.4.3 Use of the Diffusion Solution

When the diffusion equation is utilized, it is assumed to apply through-
out the entire medium including the region adjacent to a boundary. The
effect of the boundary is imposed on the solution by utilizing a jump
boundary condition.
If a real gas is considered, three coefficients must be evaluated as
given by equations (3-39), (3-49), and (3-50). However, each of these
depends only on local conditions so that they can be tabulated. Gray stagnant gas between parallel plates. -Most gases have
strong variations of properties with wavelength and it is necessary to
solve the diffusion equation in a number of wavelength regions. For
illustrative purposes here it is not feasible to consider an involved
spectral solution. There are some limited situations such as soot filled
flames and high temperature uranium gas where a gray gas approximation
can be made. The equations presented in section 3.4.2 reduce con-
siderably in this case. Let us examine then, for illustrative purposes,
the case of a gray gas contained between infinite parallel plates at
different temperatures (fig. 3-5). For simplicity let both plates have the

3-5. -Radiant interchange between infinite parallel plates enclosing gray medium.

same emissivity so that E,[,I = Elon = E,,.

For a gray gas the absorption coefficient ah is independent of wave-
length. Then the wavelength range for integration of equations (3-39),
(3-49), and (3-50) can be 0 -+ w. Letting CLA = a , which can be taken out of
the integrals, gives


Equation (3-38) reduces to

This can be integrated directly because, with no sources or sinks in the

gas, q, is a constant in this geometry. Then, with the additional assump-

tion that a does not depend on temperature and is therefore independent

of z , the result is

Evaluating equation (3-54) at z= D yields

The e b l and eb2 are in the gas at the walls. To connect these quantities
with the wall conditions, the jump boundary conditions are applied.
Differentiating equation (3-54) twice with respect to z shows that the
second derivative terms are zero in the boundary condition equations
(3-47) and (3-48). Equations (3-47) and (3-48) then become


To eliminate the unknown gas emissive powers ebl and em, add equa-
tions (3-56) and (3-57) to obtain

Then substitute eb.2 - e b l from equation (3-55) to give

or taking the reciprocal

Equation (3-59) gives the radiative energy transfer through a gray gas
layer as a function of the gas absorption coefficient, plate spacing, and
plate emissivity. It is ratioed to the difference in the black emissive


--- Eq. (3-59)
3 .8





0 .5 1.0 1.5 2.0 2.5 3.0
Optical thick~ess,K D = aD

3--6. -Validity of diffusion solution for energy transfer through gray gas between
parallel plates.

powers of the plates, which is the maximum possible energy transfer

between black plates with no intervening gas. A comparison of this
diffusion solution with the exact analytical solution of the same gray gas
problem by solution of the integral equations (ref. 4) is shown in figure
3-6. Agreement is seen to be excellent for all optical thicknesses in
this geometry.
3 . 4 . 3 . 2 The discontinuity in emissive power between two gas regions. -
Consider now two semi-infinite regions adjacent to one another (fig.
3-7). Let us determine the discontinuity in emissive power, if any, that
might occur at the interface between the regions iq the absence of
heat conduction. First, consider the media in the two regions to have no
internal heat sources or sinks. Both media are gray and stagnant; the
lower region has a constant absorption coefficient a,, and the upper has
a*. The emissive power jump at the interface between the two media is
found by integrating equation (3-52) over all wavelengths. Noting that
axl = al and ahz=az and that derivatives with respect to x and y are
zero for the one-dimensional layer being considered gives

Second derivatives with z have been taken to be zero by noting that in

either region equation (3-38) gives

3-7. - Ceornetry for derivation of interface emissive power discontinuity.


The q, must be constant since no heat sources or sinks are present.

Therefore, in either region,

Also, q, must be the same i n either region because the radiative flux
will be continuous across the interface. Therefore, equation (3-61)
can be substituted for the first derivatives in equation (3-60) to give

This reduces to

so that no discon.tinuity i n emissive power exists i n this case.


Consider now the presence of uniform volumetric energy sources of

magnitude GI and 6 2 in regions 1 and 2, respectively. Now the flux
gradient in the z direction is given in either region by

In this case, then, the second derivatives of the eo with z are obviously
not zero. Equation (3-52) then becomes

Again, equation (3-61) must hold in either region. At the interface

between the two media, since the flux is continuous,

Substituting equations (3-65) and (3-67) into equation (3-66) to

eliminate the first and second derivatives of e b gives

which reduces to

The discontinuity in emissive power is seen to exist whenever there is

energy generation in either region unless the ratios GJal and Gz/az
are equal.
For this situation, formulation in terms of integral equations (ref. 5)

The diffusion solution, although giving the correct functional dependence

of the emissive power discontinuity in terms of the a and G values,
does differ from the exact solution by a factor of 312. Other diffusion solutions for gray gases. -In table 3-11, ~ 0 1 ~ -
tions for the temperature distributions and energy transfer in simple

Geometry ~eiations~
Infinite parallel plates
1t El
' = 4~ t E ~ t E ~ t l

q(z) = C 1' 2 (D - z) t E2 t 1
ltEl[4 2

Infinitely long concentric cylinders

1t El

Concentric spheres

Q1 b= -
1 - -D l -

; a ~ ~ f 2 2 tlti)+ t + i)

p(2) ! =l + ELl is[(;, 7 I;:) $3)

- Z a ~ ~ - l - t-1.
l t ( E2 +-

a ~ e f i n i t i o nEN = 1 - v , N , , N , I= Q / , , , ~ A ~ ( -~) ], H i )= [At)- ~41($11=-


geometries involving gray gases contained between gray walls are

gathered. These equations are derived from the diffusion equations,
and caution is advised in their application since real gases are usually not
gray nor optically thick in all wavelength regions. Agreement with exact
solutions is sometimes not as good for cylindrical or spherical geometries
as for the infinite parallel plate case. Agreement has been found to be
excellent in cylindrical and spherical geometries for all parametric varia-
tions so long as the optical thickness is greater than about seven, with
better agreement as wall emissivity becomes lower and diameter ratios
(Di,lrler/Doltter) become larger. A comparison for the cylindrical geometry
will be discussed later in connection with figure 3-11.

3.4.4 Final Remarks

The diffusion solution, because of its usefulness in treating difficult
problems by standard analytical techniques, is a powerful method and is
recommended for use whenever the assumptions used in the derivation
are justified.
The most stringent assumption is that of "optically thick conditions,"
which usually is the assumption limiting application of the method.
Because most gases have line spectra, they are optically thick within the
wavelengths encompassed by the lines. Here the radiation absorption
mean free path is quite small, and the assumption that only local condi-
tions affect the spectral radiant flux is quite good. At other wavelengths
the gas can often be considered transparent and diffusion methods are
then not justified. Care must be taken, then, in applying the diffusion
equation only in geometrical and spectral regions where the assumption
of an optically thick gas is valid.
The Rosseland mean absorption coefficient should not be used as the
criterion for optical thickness. It may have a large value itself, but the
spectral absorption coefficient used in calculating a~ may be very small
in certain spectral regions. Use of the Rosseland mean in such cases
may lead to large errors. The remedy is to use wavelength bands in which
the spectral absorption coefficient is everywhere large and evaluate a
Rosseland mean for each of these regions.
Howell and Perlmutter (ref. 6) have applied the diffusion solution to
a real gas situation and have compared the results to those from an
exact formulation by the Monte Carlo method. The agreement was
generally not as good as for gray gases.

EXAMPLE3-5: The space between two diffuse-gray spheres (fig. 3-8)

is filled with an optically dense stagnant medium having constant absorp-

3-8.-Radiation across gap between concentric spheres with intervening ~ilediur~i
of constant absorption coefficient.

tion coefficient a. Compute the heat flow Q1across the gap from sphere
1 to sphere 2 and the temperature distribution T ( r ) in the gas using the
diffusion method with jump boundary conditions.
For a gray medium with constant a, equation (3-37) gives the heat
flux in the positive r direction

4 deb

The q,. varies with r according to q,.= Ql/4rr2.Substitute into equation

(3-70) and integrate from R1 to R2 to obtain

The ebl and ewz are i n the gas at the boundaries and the jump boundary
conditions are needed to express these quantities in terms of wall
values. The jump boundary conditions are given by equations (3-47)
and (3-48) and involve second derivatives which will now be found. By

integrating equation (3-71) from R , to r, there is obtained

Substitute r = (x2+ y2+z') and differentiate twice with respect to

x to obtain

Similarly for the y and z directions.

In the boundary condition (eq. (3-47)) the point 2 can be conveniently
taken on figure 3-8 at x = y = 0 and z = RZ. This gives

Also ( q , )2 = Q114rR;. Substituting into equation (3-47) gives

Similarly at the inner sphere boundary from equation (3-48)

Adding equation (3-75) and (3-76) gives

After substituting this into the right side of equation (3-72), the result
is solved for QI to give the Ji in the last entry in table 3-11.
To obtain the temperature distribution, integrate equation (3-71)
from Re to r to give

Add equation (3-75) to eliminate e M

This gives the last expression for cp in table 3-11.


Before continuing to discuss solution methods in radiative transfer,
some comments are warranted on the use of mean absorption coeffi-
cients that have been formed by an integration over all wavelengths.
Using a mean absorption coefficient avoids the need to carry out a spec-
tral analysis and then integrate over all wavelengths to obtain total
energy quantities. The question is whether it is possible to decide in
advance what mean absorption coefficient will yield an accurate solution
for a particular problem. Let us now examine in detail the mean coeffi-
cients that have been defined thus far and the relations between these

3.5.1 Some Mean Absorption Coefficients

To this point, three general types of mean absorption coefficient
have been defined. In equations (2-19) and (2-21) were defined the
Planck mean

and the incident mean

IOs an(h, T , P)ir, i(h)dh

ai (T, P ) =
1; i n , t(A)dA

In connection with the diffusion solution, the Rosseland mean absorp-

tion coefficient was defined in equation (3-39); thus,

It is instructive to examine these various mean absorption coefficients

and their relations to each other.
The incident mean absorption coefficient can be conveniently utilized
only uncles certain restrictive conditions when the incident intensity
has a spectral form that remains fixed so that the a, can be evaluated
and tabulated. For example, the situation of the incident energy being
a solar spectrum occurs sufficiently often that the a, could be tabulated
for this case. The a, is useful for the transparent gas approximation when
the spectral intensity leaving the boundaries is known, as this spectrum
will remain unchanged while traveling through the gas. If the mean
spectral intensity Ex,, is proportional to a blackbody spectrum at the
temperature of the position for which ax(A, T, P ) is evaluated, that
is, i x , , a iLb(A,T) then the incident mean becomes

ai ( T ,P ) =
1; ak(h, T, P)i&,(A, T)dh
=ap(T, P) (3-81)

In this special case only, ai becomes equal to up.

At first glance, the Rosseland mean as defined by equation (3-80)
appears to be entirely different in character from ap and ai which are
weighted by spectral distributions of energy or intensity. However, let u s
take equation (3-37) and write, for a-one-dimensional diffusion case,

But dA/dz is zero since A and z are independent variables so that, for
the difusdon case only,

Substituting equation (3-83) into equation (3-80) gives

aH ( T , P ) =

The Rosseland mean is thus seen to be an average value of ax weighted

by the local spectral energy flux dqx,, through the assumption that the
local flux depends only on the local gradient of emissive power and the
local ah.
For a gray gas the absorption coefficient is independent of wavelength,
ax(X, T , F') = cr(T, P ) , and equations (3-78) to (3-80) reduce to

as would be expected.
Determination of any of the mean coefficients from spectral absorp-
tion coefficients usually requires tedious detailed numerical integrations.
Even so, if appropriate mean values can be successfully applied to
yield reasonably accurate solutions, the time involved in solution of
many radiation problems can be considerably decreased.

3.5.2 Approximate Solutions of the Transfer Equations Using Mean

Absorption Coefficients

In this section are reviewed some of the references where mean absorp-
tion coefficients have been used in radiative transfer calculations. Solving
the transfer equations is considerably simplified when a mean absorptioil
coefficient is present or assumed because the integrations over wave-
length are not needed. This is contrasted with exact solutions for real
gases which require that these integrations be performed during each
solution. It would be inlpossible to perform all the required integrations in
advance so that they would not be needed during each calculation. For
example, the incident energy absorbed at each location depends on the
incident mean absorption coefficient a; which is weighted according to the
incident spectrum. Since this spectrum can have an infinite variety of
forms, the ai cannot be conveniently tabulated in advance. Also the
ah present in the expone~~tial attenuation terms in equation (3-1) for
example cannot be conveniently averaged over wavelength.
To avoid having to carry out spectral calculations and then integrating
over wavelength, certain approximations are often made. The simplifying
assumption most often used is that the gray gas equation of transfer
(2-28) can be used for a real gas by substituting an appropriate mean
absorption coefficient in place of the a for the gray gas. In section 2.4.2 it
was already shown that, although the Planck mean may indeed be used in
part of the energy balance equation (i.e., the term dealing with local emis-
sion), the use of the same mean coefficient in the absorption and attenua-
tion terms is invalid except in special cases. Patch (refs. 7 and 8) has
shown, by examination of 40 cases, that simple substitution of the Planck

mean in the gray gas equations leads to errors in total intensities that
varied from -43 to 881 percent from the solutions using spectral prop-
erties in the equation of transfer and then integrating the spectral solu-
tion. Reductions in error were found by dividing the intensity into two or
more spectral bands and using a Planck mean for each individual band.
In an effort to improve this situation, a number of other mean absorp-
tion coefficients have been introduced. Sampson (ref. 9) synthesizes a
coefficient that varies from the Planck mean to the Rosseland mean as
the optical depth increases along a given path. He finds agreement within
a factor of two to exact solutions for various example problems. Abu-
Romia and Tien (ref. 10) apply a weighted Rosseland mean over optically
thick portions of the spectrum, and a Planck mean over optically thin
regions and obtain relations for energy transfer between bounding sur-
faces. Planck and Rosseland mean absorption coefficients for carbon
dioxide, carbon monoxide, and water vapor are also given as an aid to
such computations.
Patch (refs. 7 and 8) defines an effective mean absorption coefficient as

[ o x a A ( ~ T, , P ) ~ ; ~ (TA) exp
, [ - - a ~ ( hT, , P ) S ] d A
a,@, T , P) =
1; &(A. T ) exp [--ah(A, T , P ) S l d h

The values of a,(S, T , P ) can be tabulated as a function of temperature

and pressure as for the other mean absorption coefficients. In addition
a, depends on the path length S and must be tabulated as a function of
this additional variable. For S small, a, approaches a,.. For very large
S, the exponential term in the integrals causes a, to approach the mini-
mum value of ax in the spectrum considered. In the radiative transfer
calculations the approximation is made that the real gas with T , P
variable along S is replaced along any path by an effective uniform gas
with absorption coefficient a,. T h e computations are then performed
using a, in the gray gas equation of transfer. The a, value used is found
by equating a,S at the T , P of the point to which S is measured, to the
optical depth of that point in the real gas. For 40 cases Patch {refs. 7
and 8) shows agreement within -25 to 28 percent of the integrated
spectral solutions, as compared with the - 43 to 881 percent agreement
using up as discussed previously in this section. This method has value
in computer-oriented solutions where the tabulated values of a,(S, T , P )
can be effectively manipulated.
Other methods of using mean coefficients are given in references 11
to 15.


In section 3.3, it was found that in certain situations it is possible
to neglect one or more terms in the equation of transfer. Solutions
using the resulting simplified equation are of course much easier than
solving the entire equation.
In this section some analytical methods are presented that account
for a,ll terms in the equation of transfer. However, only approximate
solutions of the complete equation will be sought so that, while some
accuracy may be lost in obtaining the solution, the ability will be gained
to obtain closed-form analytical solutions in many cases. This makes
it possible to gain insight into the important governing factors of the
radiative transfer, in addition to obtaining answers that are often of
acceptable accuracy.

3.6.1 The Astrophysical Approximations

As mentioned in the first chapter of this volume, much work has
been done in the study of stellar structure by analysis of the observed
radiation. Quite early in the twentieth century, astrophysicists consid-
ered the mathematical properties of the equation of transfer and applied
some approximations which remain useful today. However, these ap-
proximations were developed for one-dimensional layers of an atmos-
phere which is the case most useful in astrophysics, and the extensions
to multidimensional problems is not always obvious or possible. In
this section, two of these approximations will be examined briefly.
For more detailed treatments, see references 16 and 17. The Schuster-Schwarzschild approximation. - T h e sim-
plest approximation is to assume that, in the one-dimensional equation
of transfer, the intensity in the positive direction is isotropic and that
in the negative direction has a different value but is also isotropic.
This is illustrated in figure 3-9.
The equation of transfer is written for the intensity in each hemi-
sphere as

-- ,!3 di;,- ( p , X) =i;,-(/3, X I -iLb(x)
ah ax

These equations are now integrated over their respective hemispheres

to give


3-9.-Approximation of intensities being isotropic in positive and negative

ax dx
I d 2
i;,+ (p,x) cos p sin P dp=Io
~ / r
i;,, (p,x) sin p dp

- i;, (x) sin P dp (3-87a)

Now assume that i:, and i;,- are each independent of P ; the i t are
then taken out of the integrals and the integrations are carried out to

Equations (3-88) with appropriate boundary conditions can be solved

by use of an integrating factor as was done in section 2.3.2 for the equa-
tion of transfer. For the geometry of figure 3-5, if i ; , +(0)and i ; , -(D)are
the spectral intensities at the walls for a gas layer between planes,
the i i . + ( x ) and i [ , - ( x ) are (letting KX=UAX)

i:. + ( K A ) = i)(,+ ( 0 ) exp ( - KA) + iLb( K : ) exp ( K T - K A ) ~ K (3-89a)


From equation (2-39), the net spectral heat flux in the positive x
direction is (with the assumption that i;, + and i ; , - are isotropic)

~L,+(K) :1 cos P sin P dp+ i;,-

For the simplified case of a gray gas between parallel plates with
no internal heat sources, the q is a constant and i & , ( K ) = u T 4 ( K ) / v .
Then d q / d ~ =0 giving

Substituting equations (3-89) yields

i K Dp - + K~
I.A ( K * ) exp ( K - K * ) d ~ * - i ; ( ~ )

r i 6 (K)= 0 - T 4 ( ~ ) mi: (0) exp ( - K )

+ 0-T4(K*) exp (K- K *) d~ * ]

This is an integral equation for the temperature distribution in the gas
layer and is analogous to the exact formulation in equation (2-46).
Chandrasekhar (ref. 17) has extended this method as originally
developed by Schuster (ref. 18) and Schwarzschild (ref. 19) by divid-
ing the intensity into mean portions from discrete directions, terming
this the "discrete ordinate" method. The discrete ordinate method
has been shown to be equivalent to the differential approximation o r
moment method which will be described in section 3.6.2 (ref. 20). The Milne-Eddington approximation. -With respect to t h e
intensity, the approximation made independently by Eddington (ref. 21)
and Milne (ref. 22) is the same as that of Schuster and Schwarzschild.
It is that all the intensity traveling along all paths crossing a unit a r e a
oriented normal to the x direction with positive components in x has
a constant value with angle, and all intensity with a negative x compo-
nent has a different constant value; that is, the radiation in each direction
can be considered isotropic (fig. 3-9). However, the approximation i s
made one step later in the heat flux equation than in the Schuster-
Schwarzschild method.
Start with the one-dimensional equation of transfer (2-24). T h e n
multiply by dw and by cos P d w to obtain the two equations

cos2 p di;
dw= cos p(i;- i;,) dw
ax ax

The reason for doing this is that i; cos /3 is related to the heat flux,
and equations (3-92) will thus yield a pair of equations involving qh. If
equations (3-92) are integated over all solid angles, the result is

The assumption is now introduced that the i; is isotropic in each

hemisphere. Then

= dqx
ax dx
i;,+ :1 2 n s i n p dP+i;,-
C2277 sin p d p - 4niLb (3-94a)

qh=-(lh 1 [x
di;,+ 2n]on'2 cosz P sin P dP+- di;
dx - Id2
" 2 n cos2 p sin P d~

Integrating gives

Eliminating i;,+ i;,- between these two expressions gives

For the situation of a gray gas layer with no internal heat sources,
equation (3-97) is integrated over all wavelengths and d2q/d~"o

This is the same relation as for the diffusion approximation used to ob-
tain equation (3-54).

3.6.2 The Differential Approximation

The differential approximation reduces the integral equations of
radiative transfer in absorbing-emitting media to differential equations
by approximating the equation of transfer with a finite set of moment
equations. The moments are generated by multiplying the equation of
transfer by powers of the cosine between the coordinate direction and
.the direction of the intensity. This is a generalization of the method of
Milne and Eddington as equations (3-92a) and (3-92b) have been multi-
plied by (cos p)O and (cos O ) l , respectively. As will be discussed, t h e
first three moment equations have a definite physical significance s o
that developing a solution method by this technique has some physical
basis. The development will be given in a three-dimensional coor-
dinate system so that general geometries can be treated. The treat-
ment due to Cheng (refs. 23 and 24) will be followed here. Other per-
tinent references are 20 and 25 to 29.
A rectangular coordinate system with coordinates x l , xz, xg is shown
in figure 3-10(a). The variation in the intensity at position F along t h e
S direction in the direction of the unit vector ? is given by the equation
of transfer (2-4)

-= a ~ ( S ) f [ i(;S, ) - i: ( S ) ]
Let ah be assumed constant and integrate over all wavelengths to obtain

=a[ib(S) -il(S)]

It should be emphasized that, although the simplified notation i' (S)

is used, the intensity is a function of position and angular direction
vectors i l ( i - , ; ) as shown in figure 3-lO(a). In terms of a three-dimen-
sional coordinate system xl, x2, x3, equation (3-99) can be written a s
di' ai' a x , ai' ax2 ai' ax,
note that-=--+-
d~ ax, as -+--
axz as ax3 as

where the lj's are the direction cosines (fig. 3-10(a)), lI = cos P , / ~ = C O6,
and 1 3 = cos y.
The moments of i' are generated by multiplying i t by powers of the
l i and integrating over all solid angles. Some new notation is introduced
to designate the moments


The zeroth-order moment i'(O) has the physical significance t h a t divid-

ing it by the speed of light gives the radiation energy density as shown
by equation (2-51). The first moment i j ( ' ) is the radiative e n e r g y flux
in the j coordinate direction as shown in equation (2-52). T h e second
nloment iJfi2)divided by the speed of light can be shown to b e the radi-
ation stress and pressure tensor. The higher moments have n o specific
physical significance associated with them and are generated b y a n a l o u
with the first three.
The moment equations are obtained by nzultiplying equation (3-100)
by powers of the l i and integrating the result over all solid angles a.
The zeroth-order moment equation is the integral of e q u a t i o n (3-100)
itself, or noting that ii is independent of angle and applying the defi-
nitions of if(())and it(')

Multiplying equation (3-100) by Lk and integrating give the first-order

moment equation

9 14=
I= 1
d x , d o = ~[ i ; / , , L k d ~ - - / , ~ l J ' d ~ ]

which can be written as ( k = l , 2, 3)

This procedure is continued to generate for example the nth-order mo-

ment equation of the form

By continuing the process used to obtain equations (3-102) to (3-104),

an infinite set of moment equations can be generated as n -+ m.

'j A
Imaginary plane
in gas adjacent
, - to wall-,
*o-- --
1 'qj,t 'qj= q j , t - q j , -

(a) Coordinate system showing intensity (b) Heat fluxes in boundary condition-
a s a function of position and angle for
differential approximation.

3-10.-Differential approximation.

The next step is to approximate the infinite set of moment equations

by a finite set. When such a truncation is carried out, there will in general
be n equations in n + l unknowns. The additional equation needed
to relate the moments and provide a determinate set is obtained by rep-
resenting the unknown angular distribution of i' as a series of spheri-
cal harmonics and then truncating this series after a finite number of
terms. This whole procedure becomes quite complicated and will only
be briefly treated. It is the final differential approximation obtained
in equation (3-112) that is of most importance here.
The series expansion used to represent i' is

where A;" (7) are coefficients to be determined and

The PI)^ (cos P) are associated Legendre spherical harmonics (ref. 30),
defined by

X sin [ ( l + m+ 2k+ l ) P ] (3-107)

where I'(8) is the Gamma function, and the notation ( a ) k signifies

Equations (3-106) and (3-107) are substituted into equation (3-105)

and the resulting series is truncated by setting A/" (F) = O for 1 2 2 .
This gives an equation for i' (F, 3) which is substituted into the first
three moment equations to give

where 6 k j is the Kroneclcer delta. The form of these equations has been
considerably simplified by applying the orthogonality relations for
spherical harmonics (ref. 31). Further, note that the first moment of
i' is shown by equation (2-52) to be the energy flux, or for the j direction

Eliminating A # ( ; ) by combining equations (3-108) and (3-109) gives

Then take equation (3-111) and substitute equations (3-102), (3-103),

and (3-110) to eliminate i 1 ( O ) , and i'('), respectively. This results
after letting w T 4 = 4 in the jirst diferential approximation to the
equation of transfer ( k = 1, 2, 3 )

As mentioned by Cheng (ref. 23), equation (3-111) is equivalent to the

assumption that the radiation pressure is isotropic, which in turn is
equivalent to assumingradiative equilibrium in the gas.
The derivation briefly presented here by use of the moment equations
can be developed in a more mathematically rigorous form by use of t h e
spherical harmonic method, as was done in reference 24. The spherical
harmonic method requires considerably more algebraic manipulation
and results in the same equations developed here.
It is interesting that for a % 1, equation (3-112) reduces to the diffusion
approximation as given by equation (3-37). For a 4 1, equation (3-1 12)
reduces to

where C is a constant of integration. As pointed out by Cess (ref. 251,

equation (3-113) is the correct optically thin limit only in certain cases. Boundary conditions. -Consider a gray boundary Aj which is
perpendicular to the xj direction as shown in figure 3-10(b). The net r a d i a -
tive flux leaving A j in the positive xj direction is

where go and q, are the outgoing and incoming radiation fluxes. T h e (I,,,
however is equal to the radiation flux in the gas traveling in the negative
direction at the wall (fig. 3-10(b))

q.I , , . = qJ., - (*.J -+ 0)

The net flux in the gas in the positive x direction is qj = qj,+ - qj,- so
that qi,j = - q j ( x j 0 ) qj,+( x j -+ 0 ) . Now note that qj,+( x j -+ 0 ) is

equal to the outgoing flux from the wall q,, j so that

Substituting into equation (3-114) gives the boundary condition

The outgoing flux can also be written in terms of the intensity leaving
A j as
qo,j= l d j i ; d o (3-1 16)

where Lj is the cosine of the angle between i; and the xj direction.

A general form for the intensity is now found by substituting equations

(3-106) and (3-107) into equation (3-105) and truncating as before. The
moment equations are used to determine the A;"(?) and after consider-
able manipulation the equation for i t ( ? , s') is found to be

it ( 7 , 2) =- [it(0)(F') 39, sin 0 sin p
+ 391 cos p + 342 cos 0 sin p ] (3-117)

As a specific case, assume the boundary surface is normal to the xl

direction. Then equation (3-117) is substituted into equation (3-116) to

Qo, 1 - +
[ i t ( ' * ( i ) 3q3 sin 0 sin

+ 391 cos p + 392 cos 0 sin P] .z,-o

cos p sin p dpde

which reduces to
i1(O)(xl-+0) q, (~1'0)
(3-1 18)

4 2

where y l ( x l p O ) is the net heat flux in the xt direction in the gas adjacent
to the wall. The it(")can be eliminated by using equation (3-102) and then
i t ( ' )is eliminated by using equation (3-110). This gives

Combining equations (3-119) and (3-1 15) written for j= 1 to eliminate

go, 1 gives the boundary condition

Equations (3-112) a.nd (3-120) comprise the governing equation and

boundary condition for the differential approximation to radiative trans-
fer. Stone and Caustad (ref. 26) give a formulation for nongray gases for
the astrophysical boundary condition of zero incident flux at one
boundary. Applications of the differential approximation. - Consider
the case of infinite parallel gray plates (as in fig. 3-5) with emissivities
elcl and ~ ~ att temperatures
2 ~ T,,1 and T w z , separated by a distance D
and having a gray gas between them. The heat flux traveling through the
gas is independent of x and y and by conservation of energy is constant
with z so all dqj/dxj= 0. Then the differential equation of transfer (3-112)
reduces to

The boundary condition at z= 0 becomes from equation (3-120)

where the subscript g is used in Tf, to emphasize that this temperature

is in the gas. At z= D,

the negative sign in equation (3-123) arising because the normal direc-
tion from the surface into the gas is in the negative z direction. These
are precisely the equations found for the parallel plate case in section



Infinite parallel plates

I infinitely long concentric cylinders / 1

(2) l + t l {8[ ($ + 4 2 + i)};
- 3 a D In- 3:$] E
(2 t

I Concentric spheres I I by use of the diffusion approximation. Thus for infinite parallel

plates the predicted temperature distributions and heat transfer by
the differential approximation will be the same as for the diffusion
approximation. I n table 3-111 are also shown the differential predictions

-Exact (ref. 32)
--- Diffusion approximation
--- Differential approximation
X- .8
P .6

2 .4

.-nE .2

0 1 2 3 4 5 6 7 8 9 1 0
Optical thickness, a(Douter - Dinner)/2
3-11.-Con~parison of solutions for energy transfer between infinitely long con-
centric black cylinders enclosing gray gas.

for concentric cylinders and spheres. Comparison with table 3-11 for
the diffusion approximation with second-order slip shows that for these
geometries the results by the two methods differ only by constant
factors that have little effect on the predictions in most cases. A com-
parison with the exact numerical solution of reference 32 is shown in
figure 3-11 for the case of black concentric cylinders. Note that agree-
ment with the exact solution is better in some ranges for the diffusion
approximation, and in others for the differential approximation.


This chapter is an attempt to survey briefly the most important of
the many approximation techniques that are used for solution of the
equation of transfer. The transparent, emission, and cold material
approximations are valuable in certain simplified cases; the diffusion
solution with slip boundary conditions is easy to use and accurate when
the limitations are satisfied. The astrophysical approximations for one-
dimensional layers are of interest chiefly from a historical standpoint,
although they are still sometimes useful. The equations resulting from
the differential approximation are becoming widely used because of
their simplicity and accuracy. The gray gas equation using mean absorp-
tion coefficients is often applied, but can produce large errors in some

1. DEISSLER,R. G.: Diffitsion Approximation for Thermal Radiation in Gases with Jump
Boundary Condition. J. Heat Transfer, vol. 86, no. 2, May 1964, pp. 24e246.
2. ROSSELAND, S.: Theoretical Astrophysics; Atonlic Theory and the Analysis of Stellar
Atmospheres and Envelopes. Clarendon Press, Oxford, 1936.
3. HOWELL,JOHNR.: Radiative Interactions Between AbsorbinyEmitting and Flowing
Media with Internal Energy Generation. NASA T N D-3614, 1966.
4. HEASLET,MAX A,; A N D WARMING, ROBERTF.: Radiative Transport and Wall Tem-
perature Slip in an Absorbing Planar Medium. Int. J. Heat Mass Transfer, vol. 8.
no. 7, July 1965, pp. 97F994.
5. HOWELI,,JOHNR.: On the Radiation Slip Between AbsorbingEniitting Regions with
Heat Sources. Int. J. Heat Mass Transfer, vol. 10, no. 3, Mar. 1967, pp. 401-402.
6. HOWELL,JOHNR.; AND PERLMUTTER, MORRIS:Monte Carlo Solution of Radiant Heat
Transfer in a Nongray Nonisothermal Gas with Temperature Dependent Properties.
AIChE J., vol. 10, no. 4, June 1964, pp. 562-567.
7. PATCH,R. W.: Effective Absorption Coefficients for Radiant Energy Transport in
Nongray, Nonscattering Gases. J. Quant. Spectrosc. Radiat. Transfer, vol. 7, no. 4,
1967, pp. 611-637.
8. PATCH,R. iv.: Approximation for Radiant Energy Transport in Nongray, Nonscatter-
ing Gases. NASA TN D-4001, 1967.
9. SAMPSON,DOUGLASH.: Choice of an Appropriate Mean Absorption Coefficient for
Use in the General Grey Gas Equations. J. Quant. Spectrosc. Radiat. Transfer,
vol. 5, no. 1, 1965, pp. 211-225.
10. ABU-ROMIA,M. M.; A N D TIEN, C. L.: Appropriate Mean Absorption Coefficients
for Infrared Radiation in Gases. J. Heat Transfer, vol. 89, no. 4, Nov. 1967, pp.
11. GRANT,IANP.: On the Representation of Frequency Dependence in Non-Grey Radia-
tive Transfer. J. Quant. Spectrosc. Radiat. Transfer, vol. 5, no. l , 1965, pp. 227-243.
12. STEWART,JOHN C.: Non-Grey Radiative Transfer. J. Quant. Spectrosc. Radiat.
Transfer, vol. 4, no. 5, 1964, pp. 723-729.
13. THOMAS,M.; AND RIGDON,W. S.: A Simplified Formulation for Radiative Transfer.
AIAA J., vol. 2, no. 11, Nov. 1964, pp. 2052-2054.
14. LICK, WILBERT:Energy Transfer by Radiation and Conduction. Proceedings of the
1963 Heat Transfer and Fluid ISlechanics Institute. Anatol Roshko, Bradford Sturte-
vant, and D. R. Bartz, eds., Stanford Univ. Press, 1963, pp. 14-26.
15. HOWE,JOHNT.; AND SHEAFFER, YVONNES.: Spectral Radiative Transfer Approxima-
tions for Multiconlponent Gas Mixtures. J. Quant. Spectrosc. Radiat. Transfer.
vol. 7, no. 4, 1967, pp. 695-701.
16. KOURGANOFF, VLADIMIR:Basic Methods in Transfer Problems; Radiative Equilib-
rium and Neutron Diffusion. Dover Publications, 1963.
17. CHANDRASEKHAR, SUBRAHMANYAN: Radiative Transfer. Dover Publications, 1960.
18. SCHUSTER, A.: Radiation Through a Foggy Atmosphere. Astrophys. J., vol. 21, Jan. 1905,
pp. 1-22.
19. SCHWARZSCHILD, K.: Equilibrium of the Sun's Atmosphere. Gesell. Wiss. Gnttingen,
Nachr., Math.-Phys. Klasse. vol. 1, 1906, pp. 41-53.
20. KROOK,MAX: On the Solution of Equations of Transfer. I. Astrophys. J., vol. 122,
no. 3, Nov. 1955, pp. 488-497.
21. EDDINGTON, A. S.: The Internal Constitution of the Stars. Dover Publications, 1959.
22. MILNE, E. A.: Thermodynamics of the Stars. Handbuch der Astrophysik, vol. 3.
J. Springer, Berlin, 1930, pp. 65-255.
23. CHENG,PING:Two-Dimensional Radiating Gas Flow by a Moment Method. AIAA J.,
vol. 2, no. 9, Sept. 1964, pp. 1662-1664.

24. CHENG,PING:Dynaniics of a Radiating Gas With Application to Flow Over a Wavy

Wall. AIAA J., vol. 4, no. 2, Feb. 1966, pp. 238-245.
25. CESS, ROBERTD.: On the Differential Approximation in Radiative Transfer. Zeit. f.
Angew. Math. Phys., vol. 17,1966, pp. 776-781.
26. TONE, PETERH.; A N D GAUSTAD, JOHNE.: The Application of a Moment Method to the
Solution of Non-Gray RadiativeTransfer Problems. Astrophys. J., vol. 134, no. 2,
1961, pp. 456-468.
27. TRAUGOTT, S. C.: A Differential Approximation for Radiative Transfer with Application
to Normal Shock Structure. Proceedings of the 1963 Heat Transfer and Fluid
Mechanics Institute. Anatol Roshko, Bradford Sturtevant, and D. R. Bartz, eds.,
Stanford Univ. Press, 1963, pp. 1-13.
28. ADRIANOV, V. N.; AND POLYAK,G. L.: Differential Methods for Studying Radiant Heat
Transfer. Int. J. Heat Mass Transfer, vol. 6, no. 5, May 1963, pp. 355-362.
29. TRAUGOTT, S. C.; AND WANG,K. C.: On Differential Methods for Radiant Heat Transfer.
Int. J. Heat Mass Transfer, vol. 7, no. 2, Feb. 1964, pp. 269273.
30. ABRAMOWITZ, MILTONA.; AND STEGUN,IRENE A., EDS.: Handbook of Mathematical
Functions with Formulas, Graphs, and Mathematical Tables. Appl. Math. Ser. 55,
Nat. Bur. Standards, 1965.
31. WYLIE,CLARENCE R., JR.: Advanced Engineering Mathematics. Second ed., McGraw-
Hill'Book Co., Inc., New York, 1960.
32. PERLMUTTER, M.; AND HOWELL,J. R.: Radiant Transfer Through a Gray Gas Between
Concentric Cylinders Using Monte Carlo. J. Heat Transfer, vol. 86, no. 2, May 1964,
pp. 169-179.
Chapter 4. An l[~itroductionto the Microscopic Basis
for ~ a d i a t i o l in
l Gases and Gas Properties

In earlier chapters, the treatment of thermal radiative transfer through

absorbing, emitting, and scattering media has been chiefly from a macro-
scopic viewpoint. The atomic and molecular processes that govern the
macroscopic effects have been only briefly described in chapter 1.
Since the physical phenomena are understood on atomic and molecular
levels, much of the macroscopic material can be developed or at least
interpreted on a fundamental basis. In this chapter, the nomenclature
of some of the atomic and molecular processes is introduced and a quali-
tative discussion is given to relate these processes to the macroscopic
approach, Little is done to give methods of quantitative analysis; rather,
it is intended that for further information the reader will consult more
specialized literature, having gained some knowledge of the background
material from this chapter.
The analytical expressions are generally given in terms of an.gular
frequency R = 27rv = 2.rrclA (radlsec) rather than wavelength A or fre-
quency v. This is done because R is the most common cyclical quan-
tity employed in the literature related to this chapter and it is advisable
for the reader to become familiar with it. By use of R some of the equa-
tions have a shorter form resulting from the elimination of 271. factors.
In a few places the wave number q = l / A is employed because certain
correlations in the literature have been given in terms of this variable.

effective line or band width
Einstein coefficient for spontaneous emission
absorption coefficient
Einstein coefficient for absorption or induced en~ission
line shape parameter
speed of light in a medium
speed of light in a vacuum
diameter of colliding particles
electronic charge

oscillator strength, eq. (4-49)

Planck's constant
modified Planck's constant, h/27~
radiation intensity
Boltzmann's constant
separation constant in solution for JI
mass of molecule or nucleus
mass of electron
separation constant in solution for 4
mass of particle
number density, particles per unit volume; separation constant
in solution for $; an integer
total pressure
partial pressure; momentum of photon
energy per unit time
equilibrium distance between atoms
Rydberg constant
radial coordinate
re radius of electron orbit
ro classical electron radius
S coordinate along path of radiation
Sij integrated line absorption
T absolute temperature
7'0 reference temperature of 100 K in table 6 1 1 1
t time
v potential energy; volume
mass path length, pS
coordinates in Cartesian system
pressure broadening parameter in table @I1
"full" half-width of spectral line
average spacing between lines of absorption band
wave number
reduced mass
density of gas
time-dependent portion of T
time-dependent wave function
time-independent wave function
angular frequency
solid angle

for collisional broadening
for Doppler broadening
electron; equilibrium; emitted
energy state i or j
It'' band
n allowable particle orbits; for natural broadening
I-' projected; plioton
v dependent upon frequency
R dependent upon angular frequency
I directional quantity
+ true value, not modified by addition of induced emission
* complex conjugate


4.3.1 Bohr Model of Hydrogen Atom
Classical physics is unable to account for the line emission spectrum
of gases. To account for the line emission, Bohr in 1913 introduced
his theory of the atom, and in so doing departed radically from the
classical picture. Bohr's atom is constructed in its most simple form by
considering the hydrogen atom and making three basic postulates:
(1) An electron moves in a circular orbit without decay of energy,
and the orbit is subject to a balance of dynamic and electrostatic forces.
(2) Only stable orbits exist such that the angular momentum of the
electron is quantized; that is, the angular momentum of the electron
takes on only discrete values.
(3) The difference in energy for electrons present in different stable
orbits is equal to the energy of a photon required to produce a change
in orbit.
To write these postulates in mathematical form, consider an electron
of mass me and negative charge e in a circular orbit of radius re around a
stationary hydrogen nucleus. The Coulomb force of attraction on the elec-
tron exerted by the nucleus is e'/<, while the outward force from cen-
tripetal acceleration is mere% where Re is the orbital angular frequency
of the electron. This yields die force balance

The energy of the electron consists of potential (see eq. (4-19)) and kinetic
energy, giving
E =- - + ------- (4-2)
re 2

By use of equation (4-1) this can be written as

so that the electron energy has a reference level of zero as re b e c o ~ n e s

Since the electron is accelerating, classical electrodynamic theory
would dictate that it should radiate energy and consequently slow down
and spiral into the nucleus. To provide radiation in the form of spectral
lines, however, Bohr considered that the radiative energy loss must
occur in finite steps so that the energy given by equation (4-3) would
consist of a series of discrete levels. It was postulated that the allowable
states would be those for which the electron angular momentum is a
multiple of Planck's constant. Thus,

Equation (4-1) written for the nthorbit is combined with equation (4-4)
to eliminate Re,ll.This gives the allowable radii of the electron o r b i t s as

Equation (4-5) is used for the radius in equation (4-3) to yield the d i s c r e t e
energy states as
E,, = - -

Now consider the transition between two energy states. The difference
in energy between the jTh and ith states is obtained from equation (4-6) as

The energy of the photon required to produce a transition of the electron

between the two stable orbits i and j is equal to h a , where ft,, is the
photon angular frequency. Then equation ( 6 7 ) can be written a s

where R y is the Rydberg constant,

and has units of energy. If the transition is considered between the lowest
energy orbit (ground state, i = 1) and the highest energy orbit (j= m ) ,
it is seen that
Em-E,=Ry (4-1 0)

This is the energy required to remove the electron from the atom, and
Ry is considered to be the ionization potential for the hydrogen atom.
Equation (4-8) is found to predict exactly the frequencies of the spectral
line series of atomic hydrogen. However, for other atoms the prediction
of line frequencies is not accurate, and in many cases fails completely.
For atoms with a single electron in the outer shell, the theory can be
patched up to yield adequate results.

4.3.2 Sehrodinger Wave Equation

Because the Bohr theory is a rather curious mixture of classical and
quantum ideas, and because the predictions of the theory are not
adequate, a better formulation is required. This formulation is given
by modern quantum theory. The price we pay for the more adequate
predictions is a loss of the clear physical picture presented by the
Bohr atom.
In 1924, Louis deBroglie suggested that matter could have wave
properties associated with it in much the same way that a photon can
be assigned a mass. The momentum of a photon is given by

Then by analogy, for a particle of mass m,, and velocity v, an associated

wavelength can be found by letting m,,v= h/X giving the wavelength
associated with the particle as

The idea that a particle of matter can have an associated wavelength

seemed to be useless - however, experimental confirmation came in
the form of diffraction patterns produced by the scattering of electrons
frorn crystals. The patterns which are wave phenomena were pre-
dictable if the electrons (assumed to be particles of matter) were given
the wavelength predicted by equation (4-12).
If matter indeed has the properties of waves, then some form of
equation should predict the behavior of these waves. Where the waves
interfere constructively, we expect to find a particle and we expect
this interference to occur over relatively small regions of space. The
equgtion that is found to provide this behavior for the waves was de-
rived by Scliriidinger in 1926 and is known as the Schrodinger wave
equation. In the time-deperrclent form it is for a particle

where V is the time-dependent potential energy of the particle in the

coordinates of V2 and i is the imaginary root i = V'?i. Equation (4-13)
cannot be derived from a physical model as can the classical wave
equation in chapter 4 of volume I. Rather, the justification for this form
is that it predicts observable effects. We are left with constructing
physical models to fit the mathematical equation if we desire them,
rather than the usual process of fitting an equation to the physical
Schriidinger showed that the wave function 1V has certain boundary
conditions that are physically meaningful; it is single valued, finite,
continuous, and vanishes at infinity. When these constraints on *
observed, it is found that the solutions to equation ( 4 1 3 ) are eigen-
value-eigenfunction solutions. It is this fact that imposes quantization
on a system through the mathematics: quantization is not assumed,
but is a result of the boundary conditions on the Schriidinger equation.
The function W' has no direct physical interpretation. It corresponds
in some ways to the amplitude in the classical wave equation. A more
useful interpretation is to consider T as a probability density. However,
since 'P is in general a complex function, it is more convenient to treat
the real quantity F P * = /'P~%sthe probability density where W'* is
the complex conjugate of W'. The square of the magnitude of the wave
function / ' P I 2 then gives the probability density at any instant of finding
a particle of matter in a given location.
To satisfy the boundary conditions, it is possible to obtain a solution

to the time-dependent Schrodi~lgerequation by separation of variables

if the potential energy V does not depend on time. The separated product
has the form
V ( x , Y , 2 , t ) =+b,y , ~ ) 7 ( t ) (4-14)

Inserting this into equation (4-13) gives the two equations


where E is the separation constant.

Equation (4-16) has the solution (within an arbitrary multiplying

T =e x (- i ) =0 s ( ) + i sin ( ) (4-17)

and therefore by substituting into equation (4-14)

q = + (I,y,z ) exp (- i t)

(Note that the i in equations (4-17) and (4-18) is tlie imaginary number,
not the i t h energy state.) We now need to find $, the solution to the
time-independent form of Schrodinger's wave equation (eq. (4-15)) to
determine the complete wave function 4.
The wave equation will be considered here specifically for determining
energy of an electron around tlie nucleus of the hydrogen atom. The
potential energy of the electron is (based on zero potential energy at
r-+ m)

where F is the Coulomb force between the electron and the nucleus.
Using V from equation (4-19) results in the time-independent Schrodinger
equation (P-15) becomillg (the subscript of re is dropped for simplicity)

where the particle mass has been replaced by kc, the reduced mass of

the nucleus-electron system,

where M is the mass of the nucleus. The use of p accounts in the dy-
namics of the electron nucleus system for the slight motion of the nucleus
around the center of mass, an effect that was neglected in equation (4-1).
In spherical coordinates, equation (4-20) is

A separation of variables can be applied to obtain JI as a function of r,

0, and (o, where 8 is the cone angle,

Substituting into equation (4-21) results in the three separated equations

O sin 8 d8 sln- 8

where m, and I are separation constants which are specified as ml= 0,

-+I, rt2, . . ., 51; L=0, 1, 2, . . ., n - 1 ; and n = 1 , 2 , . . ., a.
The solution to equation (4-23) that is used is

@ = A exp (imlP) (4-26)

Equation (4-24) has solutior~sfor R in terms of Laguerre polynomials

that involve the arbitrary constant n , arrd equation ( 6 2 5 ) has solutions
for O in terms of Legendre polynomials. The solution for $ thus depends
on the three constants, n, I , and ml, each of which has discrete values.
These constants are called the quantum numbers, and they define the
possible discrete forrns of $.

If the radii are found for which the wave function has a large expecta-
tion value, these should correspond to the positions at which electrons
are found with a high probability. These radii are found by the usual
spatial averaging techniques

where the denominator is unity by virtue of $ being normalized as a

probability density function. When this integration is carried out, the
radii for various integer values of n are found to be exactly those predicted
by Bohr (eq. 4-5). It is again emphasized that the discrete values of r
are imposed by the mathenlatics of the Schriidinger equation, and not
by assun~ptionas in the Bohr theory.
Each of the linearly independent solutions for $ specifies a quantum
stote of the electron. The energy of the electron in the hydrogen atom is
found to he independent of the cluantunl nrlnlbers 1 and 171,. Thus there
are a large number of quantum states corresponding to the various
l and nzl that have the same energy. Such states are called degenerate.
By summing the number of such states that are present for a given
energy, it is found that there are 271' degenerate states per energy level
El,. (Actually, the treatment given here predicts 71' degenerate states:
the inclusion of electron spin provides the factor of 2.)
In statistical mechanics, it is assumed that every quantum state
in the atom is equally likely. Because there are 2n2 quantum states in a
given energy level El,, the factor 2n2 is called the statistical weight or
multiplicity of energy level n in the hydrogen atom. Other atoms will
have other statistical weights. Knowing the statistical weight allows
us to treat the total number of transitions per unit time occurring be-
tween two energy levels in terms of an average transition rate for all the
states in that level times the statistical weight. Detailed examination of
each degenerate state is unnecessary.


The concept of induced emission was introduced in section 1.5.5.
It was noted there that measuring the attenuation of a radiant beam
traveling through a medium gives no distinct information about induced
emission. This is because physically the induced emission combines
with the true absorption to produce an effective absorption smaller than
the true absorption. As far as radiation attenuation measurements are
concerned, the true absorption and induced emission effects cannot be

separated. Einstein (refs. 1 and 2) showed, however, that induced

emission must exist. Einstein's relatively simple arguments will now
be given, employing induced emission in the course of a derivation of
Planck's blackbody spectral distribution. Without induced emission
certain rules that are now available in statistical mechanics are also
violated, although statistical mechanics will not be discussed here.
Consider bound-bound transitions in an absorbing medium exposed
to incident radiation having spectral intensity i;). For simplicity let the
system be a collection of noninteracting atoms. Since blackbody radia-
tion is desired, let the medium be in a black isothermal enclosure at
uniform temperature-this is the condition for blackbody equilibrium
(section 2.3.2 of vol. I). 4 n atom in the medium can absorb incident
energy and thereby undergo a transition from energy state i to energy
state j. State j will consequently have a larger energy than i , or in other
words j is an "excited" state relative to i. The rate at which the transitions
from i to j occur will depend on the intensity of the incident radiation
field and the population of state i. Let ni be the number of atoms per unit
volume in state i . The Einstein coefficient Bij is now introduced. This is
defined as the probability per unit time and volume of a transition occur-
ring from state i to state j as a result of the incident energy flux per unit
solid angle and is a function only of the particular atomic system being
considered.-hen the number of transitions per unit time, considering
the effect of incident energy from all directions, is

Since the Einstein coefficient depends only on the states i and j for the
particular atomic system, it is taken out of the integral over solid angle.
The rate at which transitions will occur from the excited state j to
the initial state i depends on two factors. These factors are spon,ta,neous
emission which depends on the population nj in the excited state, and
induced emission which depends on the population n j and on the radia-
tion field intensity. Thus introduce Ajj as the probability for a transition
by spontaneous emission into a unit solid angle, and let Bji be the transi-
tion probability for induced emission. Then the rate of transitions from
j to i is

Otllrr texts inaludr or exclude various factors of 2 and .rr in t l ~ cd~finitions.S<,n~rtirnrs

tltr transitict~lrate is xrritten
as pro~tortionalto the spectral energy density ( 1 1 ~ ) do rather than the intensity.

Since on the average for a collection of randomly oriented emitting atoms

in equilibrium the spontaneous emission is isotropic, 477Aji is the prob-
ability of transition from j to i by spontaneous emission into all directions.
For a system in equilibrium, the principle of detuiled buluncing must
hold (ref. 3). This principle states that the transition rates upward and
downward between any two energy states must be equal when all transi-
tion processes are included. Using this principle, the dnldt from equa-
tions (4-28) and (4-29) are equated giving

where at equilibrium in the assumed isothermal black enclosure the

intensity becomes the blackbody intensity iA6. For blackbody equilib-
rium conditions the incident intensity is also isotropic so that

Then solving equation (4-30) for ihb ',0-ives

At thermal equilibrium the populations of the energy states are re-

lated according to the Boltzmann distribution (ref. 3). If Ei and E j are
the energies of the states, then the Boltzmann distribution gives

- exp [ (E'i&)]

where k is the Boltzmann constant. As discussed in section 1.3 and in

connection with equation (4-8), the energy difference E j - Ei is equal
to the energy of the photon either absorbed to produce the transition
from Ei to E j or emitted when there is a transition froill E j to Ei. Then,
in terms of angular frequency

so that equation (4-32) becomes


When equation (4-34) is applied to a system, the statistical weights

discussed at the end of section 4.3.2 must also be included in order to
account for all the degenerate states in each energy level.
When equation (4-34) is substituted into equation (4-31), the result is

The Planck blackbody spectral intensity is given by equation (2-llb)

of volume I as
evb 2Cl u"
i ~ : U = ~ = c : ( e c 2 r ~ /1~)o T -

which becomes the following after using Cg = hcz, Ce = hco/h-,h= 27rh,

and u= flij/27r:

Equation (4-35) has the same basic form as equation (4-36),and equating
these two expressions for i&, gives the following relations between
the Einstein coefficients (in absence of degenerate states):

Bij= Bji

Although at the time of the derivation induced emission had not been
discerned by experiment, the analysis outlined in equations (4-28) to
(4-38) gave strong evidence that it existed. If the induced emission
term of equation (4-29) is not included and the analysis is then carried
through, the resulting equation by the Einstein approach is

To make equation (4-39) conform to Planck's distribution, the ratio of

the Einstein coefficients must be, by comparison with equation (4-36)
and by use of equation (4-34),

This relation would make Aji dependent on the populations ni and nj

of the i and j states. Because the transition probabilities Aji and Bij
for a particular atomic system should depend only on the particular
states i and j and not the populations of these states, equation (4-40)
cannot be valid.
Suppose the properly specified Einstein coefficients given by equations
(4-37) and (4-38) are substituted in equation (4-39) in which induced
emission has been omitted. This will show what deviation should be
expected from Planck's distribution as a result of not accounting for
induced emission. Making this substitution gives

But (see eq. (2-13) in vol. I) this .is Wien's distribution! A comparison of
Planck's and Wien's spectral distribution in terms of wavelength is
shown in figure 2-7 of volume I , and this comparison is thus a measure
of the effect of induced emission on the spectral energy distribution.
Wien's curve is slightly below the Planck curve as a result of omitting
induced emission. It is evident that neglecting induced emission would
introduce only a small error in most cases of engineering interest.
Note that this, and indeed any, derivation of the Planck blackbody
distribution depends on the assumption of thermodynamic equilibrium.
Also, it is seen that Planck's distribution will not be obtained by the
foregoing arguments unless the existence of induced emission is


The equation of transfer was derived in section 2.3. It will now be
considered from a microscopic view by using the concepts of the previ-
ous section. Consider a beam of radiation of intensity iil traveling
through, a gas along a path S. Let the gas atoms (or molecules) be in one
of the two energy states i and j, with j being an excited state relative to
state i, that is Ej > Ei. Let the volume concentration of atoms in these
states be ni and nj, respectively. Along a path distance d S , the change in
the intensity of the beam will be governed by the energy added or lost ill
the clS interval. Neglecting scattering, the gains or losses are due to
spontaneous emission, induced emission, and absorption. By use of the

photon model discussed in section 2.7, and considering only transitions

between two energy states, the intensity added to the beam by spon-
taneous emission is
number of
rate of transitions ( p a r t i c e s ( energy )
particle- solid angle volume transition

Similar relations are derived for induced emission and absorption. The
equation of transfer becomes

This can be arranged into

!$3+ B.. .
a= B . . n" . ~1 n i j
[Bij ni (Bijni ) i"It
Although the system here is not in blackbody equilibrium, the Einstein
coefficients can be used as previously obtained as they depend only on
the energy states and particular atomic system being considered. From
equation (4-35) (noting that Bij = Bji from eq. (4-37))

Aji- ihb(e*"../*T-

Substituting this and equations (4-34) and (4-37) into equation (4-44)

which simplifies to

Noting that hR/kT=hc,/kAT, the quantity multiplying the i;l on the

right side of equation ( 4 4 5 ) is compared with equation (1-25). The true
absorption coefficient is thus found to be

and the absorption coefficient including induced emission is

Equation (445) becomes

This is the form given for the equation of transfer by the macroscopic
derivation (eq. (2-4)).
Thus the equation of transfer and, earlier, the form of Planck's
distribution have been derived from consideration of microscopic
processes. The true absorption coeffcient a;t is shown by equation
(4-46) to be directly related to the Einstein coefficient Bij. Rather than
using the Einstein coefficient, it is customary to give the transition rate
between bound electronic energy states in terms of a parameter called
the oscillator strength or f-number. This is related to Bij by

f.. = hntecofljj Bij


where me is the mass of the electron and e is the electronic charge.

Substituting equation ( 4 4 9 ) into equation ( 6 4 6 ) gives ah in terms of
the osciIlator strength as

where r, is the classical electron radius (see table I in the appendix).

Froin equation (4-50), it is seen that the true absorption coefficient is
directly proportional to two factors. These are ni, the population of the
initial state of the absorbing species, andfj, which through its connection
with Bij, is related to the probability per unit time for transitions to occur
from state i to j. Calculation of the population ni, at least in the case of
local thermodynamic equilibrium, is a problem in statistical mechanics.
It is possible to derive f-numbers for many electronic transitions by
quantum mechanics and thus derive ah from first principles.
The determination of the spectral absorption coefficient by means of
statistical and quantum mechanics requires a knowledge of t h e transition

processes that can occur. In complicated atoms and molecules so many

transitions are possible that calculations must either be restricted to
only the important transitions, or else statistical or simplified models
must be tried. Some discussion of the form of the spectral absorption
coefficients for various types of trarlsitiorls is given in the ncxt section.


A gas can absorb energy by a variety of nlicroscopic mechanisms.
Each of these mechanisms involves adding the energy of the absorbed
photon to the internal energy of a gas atom or molecule. A preliminary
discussion of the types of absorption processes was given in section 1.3.

4.6.1 Spectral Line Broadening

If the gas is not dissociated or ionized, then the internal energy (not
including translational energy) of the gas is contained i n discrete vibra-
tional, rotational, and electronic energy states of its atoms or molecules.
The absorption of a photon can cause a transition of some state of the
atom or molecule to a state of higher energy. Because only discrete
energy states are involved in these transitions, photons of only certain
energies can be absorbed. If the energies of the upper and lower discrete
states are E j and Ei, respectively, then only photons of energy E,,=Ej-Ei
can cause a transition. As discussed in sections 1.3 and 4.3.1, the energy
of a photon is related to its frequency through the relation

Consequently the discrete transitions result in the absorption of photons

of only very definite frequencies causing the appearance of dark lines in
the absorption spectrum. Hence this process is termed line absorption.
Because both the initial and final states of the atom or molecule are dis-
crete bound states, these energy changes between states are called
bound-bound transitions. The rates at which these transitions occur are
available in tabular form for some lnolecules and atoms (refs. 4 and 5).
The relations for the transition rates are often given by the semi-classical
results describing radiating atoms multiplied by a modifying factor,
called the Gaunt factor, that provides the correction for quantum
nlechanical effects.
Equation (4-51) would predict that very little energy could be absorbed
from the entire incident spectrum by any given absorption line, because
only those photons having a single frequency could be absorbed. Other
effects, however, cause the line to be broadened and consequently have a

finite frequency span around the transition frequency R, of equation

(4-51). The frequency span of the broadened spectral range, and the
variation within it of the absorption ability, depends on the physical
lnechanism causing the broadening of the spectral line. Some of the im-
portant line broadening mechanisms are called natural broadening,
Doppler broadening, collision broadening, and Stark broadening. For
most engineering conditions involving infrared radiation, collision broad-
ening is the nlost important.
The variation within the broadened spectral line of the absorption
coefficient with frequency is called the shape of the spectral line. These
shapes are important as they are related to the basic trends of the gas
absorption with temperature, pressure, and path length through the gas.
The shape of a typical spectral line is illustrated by figure 4-l(a). The
~ n , , ~ ( fisl )the variation of absorption coefficient within the line broad-
ened about the frequency R,, which is the tramition frequency obtained
from equation (4-51). The integrated absorption coeficient S , for a single
line is found as the integral under the entire a n , , ( R ) curve

The cln, i j ( R ) will be essentially zero except for fl close to Rij. The
regions away from Rij, where a n , i j becomes small, are called the "wings"
of the line. The magnitude of Sij will depend on the n u n h e r of nlolecules
in energy level i and hence will depend on the gas density.
The line shape parameter is defined as

so that the S i j is used as a normaliziilg factor. If equation (4-53) is inte-

grated over the range 0 s Cl s w , then substitution of equation (4-52)
shows that b i i ( R ) is nornlalized such that

By dividing bij(Rij) the line shape parameter can be given a s a function

extending from 0 to 1 as in figure 4-l(b). Note that from the definitions
there is the simple equality

-= an,ij(R)
bij ( f i t j ) an,ij (Rij)

Angular frequency, R

Angular frequency, R

(a) Absorption coefficient.

(b) Normalized line shape paratneter.
4-1.-Broadened spectral line for transition between energy levels i and j.

The form of the spectral line depends on the governing line broadening
phenomenon. One characteristic of the line shape is often expressed
in terms of a parameter A called the "jiuLL" half-zuidth of the line. The
A is the width of the line (in units of angular frequency for the present
discussion) evaluated at half the maximum line height as shown in figure

4-1.' This provides a definite width dimension to help describe the line.
Since an,,, goes to zero asymptotically as )Q- R,I increases, it is not
possible to define a line width in terms of frequencies at which an,,
becomes zero.
Four phenomena that cause line broadening will now be discussed
along with the resulting line shapes. Natural broadening.-A perfectly stationary emitter unper-
turbed by all external effects is observed to emit energy over a finite
spectral interval about a single transition freqnency. This n a t ~ ~ r line
broadening results from the uncertainty in the exact levels E, and E,
of the transition energy states, which is related to the Heisenberg uncer-
tainty principle. Tl-te natural line broadening produces a line shape
parameter of the form


where A,, is the "full" half-width of the line for natural broadening.
This form of bij is callkd a resonance or Lorentz profile. In units of
frequency, it provides a profile that is sy~llmetricabout Rij and that
depends on A,, and the transition frequency Rij. Equation (4-55) con-
forms with the various definitions imposed on it. When R = Rij, the
maximum bij is bij(Rij)=2/rA,, so that (112)bij(Rij)is ~ / v A , , . This is
the bij obtained when R - Rij is set equal to A,,/2, as expected from the
definition of A,,. The integral

dfl= I

in conformity wit11 equation (4-54).

For engineering applications the half-width produced by natural broad-
ening is usually quite small compared with that caused by other line
broadening mechanisms. Natural line broadening is therefore usually
neglected. Doppler broadening.-The atoms or molecules of an absorb-
ing or emitting gas are not stationary, but have a distribution of velocities

T l i ~I ~
'Sometimes a q ~ ~ a n t i ct qy ~ t a tl o A12 is usrd trlrirh ~ o 1 1 1 d c a l l ~ dlwre if "Italf' I~alf-~)i<Itll. B ~ P I
1111191llr varrfi~l
to I,? sure \ v l ~ i r lli~ic
~ wilftli isl),xin: usrd as thtl trrtninol<,py viirie~in thr litrraturr.

associated with their thermal energy. If an atom or molecule is emitting

at the frequency Rij and at the same time is moving at velocity v toward
an observer, the waves will arrive at the observer at an increased fre-
quency R given by

If the emitter is moving away from the observer, the v will be negative
and the observed frequency will be less than atj.An example of such a
frequency decrease is the "red shift" of the radiation detected from
galaxies in the universe. This provides evidence that the galaxies are
moving away from Earth thereby indicating that the universe is
In thermal equilibrium the gas molecules will have a Maxwell-
Boltzmann distribution of velocities. If an observer is detecting radiation
along one coordinate direction, then the velocities of interest are those
along the single direction either toward or away from the observer. The
fraction of molecules moving in that direction within a velocity range
between v and v dv is
"= JM2n-kT exp (-%)
Mu' dv

where M is the mass of a molecule of the radiating gas and k is the Boltz-
mann constant. Using equation (4-56) in equation (4-57) to eliminate v
gives the fractional number of molecules providing radiation in each
differential frequency interval as a result of Doppler broadening. The
result is a spectral line shape having a Gaussian distribution; that is,

bij(0) =-- exp [-4(0-flij)'(:)] (4-58)

where AD is the "full" half-width of the line for Doppler broadening. The
line shape parameter bij(0) depends only on AD and the transition fre-
quency Oij. The Doppler "full" half-width, however, is given by

thus depending on fitj, T, and M . The dependency of AD on T1I2shows

that Doppler broadening is important at high temperatures. Collision broadening.-As the pressure of a gas is increased,

the collision rate experienced by any given atom or molecule of the gas
is also increased. The collisions can perturb the energy states of the
atoms or molecules resulting in collision broadening of the spectral
lines. For noncharged particles, the line takes on a Lorentz profile
(ref. 4); that is,

which is the same shape as for natural broadening.

The collision "full" half-width A, is determined by the collision rate
and an approximate value can be found from kinetic theory. The A, is
given by


4-2.- Line shape parameter for Doppler and Lorentz broadened spertral lines
(areas under two curves are equal).

where D is the diameter of the atoms or molecules and P is the gas

pressure for the single component gas. Equation (4-61) shows that
collision broadening becomes important at high pressures and low
Collision broadening is often the chief contributor to line broadening
for engineering infrared conditions, and the other line broadening
mechanisms can usually be neglected. The shapes of the Doppler and
Lorentz broadened lines are compared in figure 4-2 for the same half-
width and area under the curves. The Lorentz profile is lower at the
line center, but remains of appreciable size further out in the wings of
the line than the Doppler profile. Even when Doppler broadening is
dominant near the line center, collision broadening is often the important
mechanism far from the center. Stark broadening. -When strong electrical fields are present,
the energy levels of the radiating gas particles can be greatly perturbed.
This is the Stark effect and can result in very large line broadening. It
is often observed in ionized gases where radiating particle interactions
with the electrons and protons give large Stark broadening effects. Calcu-
lation of the line shapes must be approached through quantum mechanics
and the resulting line shapes are quite unsymmetrical and complicated.
Stark and collision broadening are often lumped under the general
heading of "pressure broadening." Both effects depend on the pressure of
the broadening component of the gas. When two or more broadening
effects simultaneously contribute to the line broadening, calculation of
the resulting line shape becomes more difficult. References 4, 6, 7,
and 8 can be consulted for additional information.
Broadening has been discussed here under the assun~ptionthat only
one atomic or nlolecular species is present in the gas. If the gas consists
of more than one component, then collision broadening in the radiation
absorbing gas is caused by collisions with like molecules (self-broadening)
and by collisions with other species. Both collision processes must be
included in calculating line shapes.

4.6.2 Absorption or Emission by a Spectral Line

By integrating equation (1-37) over the entire spectrum, the total
energy absorbed along a path S per unit solid angle and projected area
can be found within a uniform gas. As shown in figure 1-10, this is the
energy absorbed when radiation with an incident intensity ii(0) travels
through the shaded solid angle and reaches dA,

= ( 0 1 e x p ( a d ) ]dR

where i&(O) is the incident spectral intensity at the origin of path S .

Similary, from the analogous forms of equations (1-41) and (1-43), the
energy can be found that is emitted to dA,, by a uniform gas in the region
of solid angle dm and length S of figure 1-10. Per unit solid angle and
per unit of the projected area, this is

lo ibu [1- exp ( a d )] d R

The integrals in equations (4-62) can be evaluated for an variations

corresponding to a broadened spectral line, but first some simplifications
can be made. Consider a spectral line having the transition frequency
flu. The line absorption coefficient an, i j ( R ) will be essentially zero
except in a narrow frequency range surrounding Rij. Hence unless S is
large the integrands in equations (4-62) will be of appreciable magnitude
only within this narrow frequency region and the integration need be
performed only over this narrow range. Within this range the iA(0) or iA,
can be approximated as being constant, and since the largest absorption
is at R i j , the i h ( 0 ) and i;lb are ordinarily taken at that frequency. Then
equations (4-62) become for the spectral line

d2Q' = iA(0,Rij) (1- exp [- ~ n , i j ( R ) S ] } d R (4-63a)


The absorbed and emitted energies for the line thus both involve the
same integral. This integral will be called the effective line width Kij
so that
(1 - exp [-an,ij(R)SI}dfi (4-64)

The Aj is a function of path length S , and has units of the spectral

variable which is R in this instance. By considering a spectral line within
~vhichthe gas is perfectly absorbing ( a n ,ij+ m), and having n o absorp-
tion outside this line, it is found from equation (4-64) that can be
interpreted as the width of a black line centered about R i j that produces
the same emission as the actual line.

The evaluation of Aj will now be considered for two important limiting

cases. First consider the situation where the optical path length a n , i j ( R ) S
is small ( a n ,i j ( R ) S< 1 ) . The exponential term in equation (4-64) can
be expanded in a series yielding

(an· ·SF
1- exp (- an , ijS) = an, IjS - 2'1 + ..

and then only th e first te rm r e tain e d. By using e quati on (4-64), the Aij
beco mes

Aij(S) = S L" an , ijdD

By use of equation (4-52) thi s yields

Aij(S) = SSij (4- 65)

where Sij i the integrated absorption coefficie nt and sh ould not be

co nfu sed with S which is th e path le n gth. Th e effective lin e width is
thus lin ear with path le ngth S in the limit when an, ij(D)S ~ 1 regardless
of the line s hape . A lin e with this lin ear be havior is called a weak line.

Next to be co nsidered is the situ ation where the optical path an, u(D)S
is large. Thi s will be done for the Lore ntz line s hape equation (4- 60)
for colli sion broad e ning, as thi s is the mos t important type of broade nin g
for e ngin eering applications in the infrared region . From equati on (4-53)
the line absorption coe ffi cie nt as a function of frequency D is

where bij (D ) is the line s hape parameter. Usin g the Lor entz line s ha pe
eq uation (4-60) for bij gives


ow s ub titute this into equati on (4-64) to ob tain Aij for th e spectral

lin e as

- f "" {
[ -2 D. 2
D. c
7T" ~+ (D-D .. )2
II dD (4- 67)

For a Jine that is very trongly absorbin g at its ce nter, th e collision "full "
half-w idth D. c is s mall and ca n be neglected co mpared with I D - D ij I
except in the s mall region where D is very close to D ij. In th e r egion
where D i clo e to Dij, th e ex ponential term in th e integrand i mall and
he n ce its accuracy is not important. As a res ult for a s trong lin e, e quation


(4-67) can be a pproxim ated as


Since it is a sin gle lin e that is bein g co nside red in equati on (4- 68), the
integrand beco mes very s mall as In - n ij I beco me large. F or a Lore ntz
profil e the lin e s ha pe is symm e tri c about n ij so that the integral can be
writte n as


To carry out th e integral, le t the variable y be de fin e d as

The n equation (4-69) can be writte n as

- _ (Sijfl cS)I /2 f "" 1 -exp ( -y)

A ij( S ) - 2 3/ 2 dy
'7T 0 Y

Thi s can be integrated to giv e

(4- 71 )

Equation (4- 71) s how that, for a strong Lorentz line, the a bsorptan ce
varies as the squa re root of the path le ngth. Thi s i in co ntras t to the
res ult for an y weak line, equa tion (4- 65), wh ere the ab orptan ce varies
linearly with path le ngth. Experim e ntal res ult bear out th ese fun cti onal
de pe nd encies.

4.6.3 Continuum Absorption

Certain ene rgy tra nsiti on processes can res ult 111 the a b orption of
photon hav in g a wide r an ge of e nergie a opposed to the relati vely s mall
range of e nergies that line ab sorpti on can e nco mpass . Th e continuum
abso rption proce ses ca n be divid ed into two catego ries, boundj ree
processes and free jree processes. Th ese processe were previou ly di s·
cusse d in secti on 1.3 and ar e briefl y r eviewed her e. Continuum ab orp·
ti on can also re ult from solid parti cles upended in the ga which will
be di sc ussed in c hapter 9.
4.6.3. 1 Boundjree processes. - Co nsid er whe n a molec ule ab orbs
a photon of s ufficient energy to c ause dissociation or ionization. A ph o-
ton of any energy greater t ha n the minimum nece ary for th ese proc-

esses can be absorbed, giving rise to a continuous absorption spectrum.

To produce ionization, ejection of an electron from a bound to afiee state
occurs upon absorption of a photon. Free-free processes.-A photon can be absorbed by a free
electron as a result of an interaction of the electron with the electric
field that exists in the vicinity of a positive ion. The energy of the ab-
sorbed photon is added to the kinetic energy of the absorbing electron,
which remains in the free state. Since the initial and final states are not
quantized, a continuous absorption spectrum results.

4.6.4 Band Absorption Correlations

The gases that are commonly encountered in engineering calculations
are diatomic or polyatomic, and therefore possess vibrational and rota-
tional energy states that are absent in monatomic gases. The transitions
between the vibrational and rotational states usually provide the main
contribution to the absorption coefficie~ltin the important thermal radia-
tion regions of the spectrum at moderate temperatures. As the tempera-
ture is raised, dissociation, electron transitions, and ionization become
more probable, and the contributions of these additional processes to
the absorption coefficient must be included.
When the absorption coefficient of a gas is determined experimen-
tally, the contributions of all the line and continuum processes are super-
imposed. In conlputing such coefficients, each absorption process
must be analyzed and then the conlplete coefficient obtained by com-
bining the contributions from the various processes. In figure 4-3, the
contributions to the spectral absorption coefficient as given by reference
9 are shown for air at a pressure of 1 atm and a range of temperatures.
The ordinate is the fractional contribution of any of the transitions to
the entire absorption process; hence, the ordinates of all curves sum
to unity at each temperature. At low temperatures the entire absorption
results from transitions of the oxygen between molecular states. As the
temperature is increased, there is some formation of NO which provides
additional bound-bound transitions. At high temperatures the continuous
absorption processes discussed in sections 4.6.3 and 1.3 are dominating.
These are bound-free (photodissociation) and free-free transitions.
The vibration-rotation bands are usually the most important absorbing
and emitting spectral regions in engineering radiation calculations.
The structure of such a band will now be examined in more detail and
this will reveal the difficulty of computing band absorption coefficients
from basic principles. Some of the simplified models of a band will then
be discussed by means of which some band absorption features can be
analyzed. The correlation of experimental band absorptance data will

1.0 Band system



s .6

0 2000 4000 6000 8000 10 000 12 000

Temperature, T, K
4-3. -Contributions of energy transitions in bands of various species to absorption
coefficient of air at 1 at111 (ref. 9).

then be considered to show how gas properties can be presented in a

manner that is useful for engineering applications where band radiation
c~uantitiesare required. Often in engineering heat-transfer problems, a
reasonable approximation to the total radiation will suffice. It is then not
necessary to go into the details of the radiation from the individual bands.
For total radiation calculations, charts of gas total emittance have been
developed from total radiation measurements. These charts will be dis-
cussed in chapter 5. Many of the functional dependencies of these charts
had been developed empirically before the details of the radiation from
the individual bands had been found. The information in the following
sections will aid in understanding from a microscopic viewpoint how
the physical variables influence gas radiation but is not intended to
yield analytical predictions of the properties.

Transition Band
0 00'0 - 0110 667.30
01'0 - 10'0 720.86
0110 - 0 2 ~ 0 667.92
0 2 ~ 0- 03'0 668.61
0 1 0 ~ 0 0 - l l ~ 0688.75
0 2 ~ 0 11'0 741.69

14.' 20 14: 15 14.'10

Wavelength, h, p m

700 705 710

Wave number, TJ, cm-I

4-4.-A portion of hiph-resolution spectrum of carbon dioxide (ref. 10)

Let us now examine in more detail the vibration-rotation transitions

governing the absorption coefficient of most polyatomic gases up to a
temperature of about 3000 K. These transitions are strong functions of
frequency, and consequently the absorption coeficient is also strongly
spectrally dependent. The spectral absorption in a vibration-rotation
band consists of groups of very closely spaced spectral lines resulting
from transitions between vibrational and rotational energy states. An
example is shown in figure 4-4 for a portion of the carbon dioxide spec-

t r u n ~ .The
~ absorption lines are so closely spaced in certain spectral
regions that the individual lines are not resolved in most instances by
experimental measurements. The lines appear to, or actually do, over-
lap as a consequence of broadening, and merge to form absorption bands.
An example of absorption bands observed with low resolution are those
for carbon dioxide that were shown in figure 1-2.
The large number of possible energy transitions that can produce
an array of spectral lines as in figure 4-4 is illustrated by the many
energy levels and transition arrows in figure 4-5. This figure shows
the potential energy for a diatomic molecule as a function of the sepa-
ration distance between its two atoms. The two curves are each for a
different electronic energy state where the electron may be shared by
the two atoms. The distance R e is the mean interatomic distance cor-
responding to each of the electronic states. The long-dashed horizontal
lines denote vibrational energy levels, while the short-dashed lines
are rotational states superimposed on the vibrational states. Transitions
between rotational levels of the same vibrational state involve small
values of Ej-Ei. Hence from equation (4-33) these transitions give
lines in band structures located at low frequencies; that is, in the far
infrared. Transitions between rotational levels in different vibrational
states give vibration-rotation bands at frequencies in the near infrared.
If transitions occur from a rotational level of an electronic and vibrational
state to a rotational level in a different electronic and vibrational state,
then large E j - E i are involved and a band system can be formed in
the high frequency visible and ultraviolet regions of the spectrum.
In a detailed radiation exchange calculation the absorbed and emitted
energy will be needed in each band region, for example in the four main
COz bands of figure 1-2. These spectral bands are separated by spec-
tral regions that are nearly transparent. A possible approach to corre-
lating gas properties is to examine the absorption of each band separately
and develop empirical correlations describing the behavior of each band.
If the absorptance of the individual bands can be correlated in terms of
the pressure, temperature, and path length through the gas, then energy
interchange methods can be applied on a band-by-band basis to compute
the total energy transfer through a real gas.
Ecluatiolls (4-62) for absorbed and ernitted energy both involve the
same type of integral, the only difference being that the integral for
energy absorbed contains the incident intensity, while that for the emitted
energy contains the blackbody intensity. Since the absorption bands

''l'lrr notntioil (01'0). figure 4-4 ih a drsignatioo used to shoxv the quantum state of a Ilarmonic oscillator. In
tlie ~r,nrralwtsr it~~trj~,.i!, IIIP I , , a r the s I is the quantunl number for angular monwrltum.
~ vil~rationalquaaturn ~ l u r n b ~ rand
'l'ransitions brtat.en Iwn Fnergy states, soclr as t l ~ o s rdenoted by (00°0)--t (01'0) give rise to absorption lines. Certain
r c ~ l w f i n ~rult,s
i gclvern the nllo~rnbletransitions. A gorrd intradt~rtory trratnlent is given in cl~nptrr3 of Goodp (ref. 10).

--Vibrational state
----- Rotational state

\ For electronic

,- Transition between rotational

/' levels i n different electronic states
Y- For electronic state 1

state 1

states of same
electronic state

Internuclear separation distance

4-5. -Potential energy diagram and transitions for a diatomic molecule.

usually occupy a rather narrow spectral region, an average value of the

i ; t ( O ) or i;lb can be taken out of the integral for each band. Consider-
ing for example the total emitted energy, equation (4-62b) becomes

where the subscript 1 denotes a band, the integral is over each band,
and the summation is over all the bands.
In similar fashion to the effective line width in equation (4-64) the
integral in equation (4-72) is defined as the effective band zoidth or

2,( S ) = 1
b n ~ t dtoidtli
(1- exp [- an (R)SlI d f i (4-73)

The 2,will have units of the spectral variable, which is R in the case
of equation (4-73). More often in the literature k, is tabulated in terms
of wave number so that it has units of cm-l. The span of the absorption
band that provides the upper and lower limits of the integral in equation
(4-73) does not have a specific value that applies for all conditions. It
can be defined as the spectral interval beyond which there is only a
given small fractional contribution to 2,. The width of this interval will
increase slowly with path length as a result of proportionately more
absorption taking place in the wings of the band.
By comparing equations (4-73) and (4-64) it is found that the A /for
the band is the sum of the Alj for all the spectral lines that occupy the
band if all the 2,. act independently of each other. Generally, the spec-
tral lines do overlap and as a consequence each line does not absorb as
much energy as if it acted independently of adjacent lines.
As was observed in figure 4-4 an absorption band is typically com-
posed of many broadened absorption lines. Hence the a n ( R ) in equa-
tion (4-73) is a complicated irregular function of frequency, and the
integration for 4 is difficult mathematically. The integration would also
require that the detailed shape of all the broadened lines be known. It
i s evident that a simplified model for the form of an(a) must be devised
if integration over the lines to obtain band radiation properties is to be a
fruitful analytical approach. Two common models are used which rep-
resent the extremes in specifying the individual line spacings and
Elsasser (ref. 11) has modeled the lines as all having the same Lorentz
shape, equation (4-66), and being of equal heights and equal spacings.
This gives an as a periodic function of R as shown in figure 4-6(a). The
periodic function depends on the parameters governing the shape of
the Lorentz line as well as on the spacing 6 between them. The absorption
coefficient at a particular frequency is found by sunlming the con-
tributions from all the adjacent lines. The distance of the line centers
from a position R are 10- 01, J R - 61, IR-261, and so forth. Then
summing all the contributions by use of the Lorentz shape in equation
(4-66) gives (each line has integrated absorption coefficient S,)

This periodic function is inserted into equation (4-73) and after a sim-
plifying transformation, the integral can be carried out numerically.
Some of the results are given in reference 10. Analytical relations can

.- *
E -26 -6 0 6 26
a Frequency, R

Frequency, R

(a) Equally spaced Lorentz lines.

(b) Statistical model.
FIGURE4-6. -RIodels of absorption lines forming an absorption band.

be obtained at the weak and strong absorption limits. These results

have been used as a guide for experimental correlations and are discussed
by Plass (ref. 12). Presenting the specific analytical results is more
detailed than the treatment intended here so the reader is referred to
references 10, 12, or 13 for further results.
Another band model is a statistical array of lines a s shown in figure
4-6(b) and presented by Goody (ref. 10). There can be either a random
spacing of identical lines or more generally the lines can also differ from
each other. A spacing that is essentially random is typical of bands of
polyatomic n~oleculessuch as C O z and water vapor. To apply the model,
probability distributions of line strengths and positions must be assumed.
These statistical assumptions remove the necessity of calculating the
exact properties of the individual lines in the band.
Many other band models have been proposed, some of them of more
utility in certain cases than the Elsasser or Goody models. Several of
these band models are discussed by Goody (ref. 10) and Edwards and
Menard (ref. 14). Modifications to the Elsasser model have been made
recently by Kyle (ref. 15) and Golden (refs. 16 and 17) who treated evenly

spaced lines with a Doppler profile, and Golden (ref. 18) who treated
the same case with a Voigt profile. The Voigt profile is a coinbination of
the Doppler and Lorentz profiles, thus accounting for the presence in
the gas of the both Doppler and collision broadening.
Once the line structure is specified that constitutes the band, the effec-
tive band width Al can be calculated from equation (4-73). It is evident
that the A; will depend on the line spacing, the line half-width and the
line integrated absorption, as well as other quantities when the random
statistical model is used. To utilize these analytical results for radiative
calculations involving a real gas mixture, it is necessary to know how
all these factors are influenced by conditions such as the gas temperature,
partial pressure of the absorbing gas, and the total pressure of the gas
mixture. If the relations between these quantities are specified, then
the correlation of experimental data based on the theoretically indicated
dependencies of the band integration can be attempted. The background
for these calculations is well charted by Goody (ref. 10). A few of the
major functional relations indicated by the theory are examined in the
remainder of this chapter. Edwards and coworkers (refs. 19 to 27) have
assembled a large body of data on the common gases and have obtained
correlations for many of the important band structures. References to
these and other data are summarized in table &I and some of the results
will be given later.
The radiative behavior of the absorption bands can be conveniently
presented by correlating the effective band width Al for the various
bands as a function of the path length, pressure, temperature, and so
forth. As will be shown in section 5.7 the A/ can be used in equation
(5-74) to obtain the band absorptance for use in detailed spectral ex-
change calculations in enclosures. By using equation (1-43) the Ar
can also be used to calculate the total emittance of a uniform gas as

( T ,P , S ) = -
rib6 [1- exp ( - a n S ) ] d o

7 'Ti-
iib,,L [l-exp(--an~)ldo

'Ti- -
aT4 I
ihb,l A1

The E' can be used, as described in section 5.6.1, for engineering cal-
culations of radiation from an isothermal gas to an enclosure boundary.

ABLE -!-~~.-~.\~AII.ABI,E

Bands 1 ~etermc~ / Ceili~lienti Type of correlatio

coy Equivalent band


Exponential wide

9.4 and 10.4pni

1 1 a22, 29 300 < T < 1400 K,
0.1 < X s 23 000 g/r/m2
Exponential wide

All important
I 31 to 34
I T - 300 K Equivalent hand

Exponential wide

1 1

2.7 and 6.3 p n ~ 26 300 G T < 1100 K, Equivalent line

1 s X c 21 000 g/ll12

All important
I 3 1 to 34
I T - 300 K Equivalent band

Exponential wide

Exponential wide

Not correlated - pre-

sented in terms of
spectral emittance

Not correlated- pre-

sented in terms of
spectral and total

Discussion of literature
up to 1960

Air All important 4 References to literature

contributing (Tal~le up to 1965 for needed
hands 11-2) data to calculate band

.'Ct,rrclations g i w o in table. 4-11 and 6 1 1 1 .

" N2, 0 1 ,C 0 2 . O , , H 2 0 . CH., and nitrogen oxidrs.

Let us nour exan~inethe behavior of A, for the limiting cases of weak

and strong absorption which will provide some liillits on which to base
band correlations. For a single spectral line, it was found in equation
(4-65) for a weak line and in equation (4-71) for a strong Lorentz line
that the effective line width varied respectively as a linear and as a
square root function of the product of path length S and integrated
absorption coefficient Sfj.If it is assumed as a first approxi~nationthat
within a band the effect of line overlap is small, then these trends would
also apply to the effective band width and can be used as a basis for a
first approximation in the correlation of experimental data. It follows
that bands composed entirely of either weak lines or strong lines must
be each correlated with a different path-length dependence. For weak
lines, the use of equation (4-65) indicates that the correlation of the
effective band width is expected to be for the 1"' band,

For a strong collision broadened band, equation (4-71) indicates that

A/ (S) (SlAcS) '1'

Considering a single gas, the integrated band absorption coefficient S1

depends on the number of molecules or atoms undergoing transitions
and hence a s a rough approximation is taken as proportional to the gas
density. The collision "full" half-width A, is given by equation (4-61)
and because of the direct dependence on pressure the A,. is taken as
being proportional to the density of the absorbing species. (The T-lI2
in A,. would also have some influence.) Inserting these dependencies
into equations (4-75) gives the weak and strong band approximations in
terms of density and path length as

Al (s) a pS (weak band)

(s) a pSl/2 (strong band)

when the lines in the bands are all acting independently.

For a very strong band absorption, that is, for very long paths and
many strong overlapping lines, the proportionality of equation (4-76b)
does not hold. It is evident that with increasing S, the A, in equation
(4-76b) would exceed its upper limit which from equation (4-73) would
be the actual width of the absorption band. For very strong absorption
there is some justification for taking the dependence of on ,a and S
to be of the form

2,(S) OC In pS (4- 77)

as discussed by Edwards and Menard (ref. 14).

A difficulty in the use of equations (4-75) and (4-76) is to know when
the correlation for a given band will go from the linear to the square-root
behavior. For intermediate absorption strengths these two limiting
regimes do not join smoothly and an abrupt transition is not physically
meaningful. An improvement was made by Edwards and Menard (ref.
14) who introduced a simplified correlation that transitions smoothly
between the regions of differing dependency on p and S. The method has
been used to successfully correlate experimental data (refs. 21 to 24).
The basis of this correlation is the assumpti011 that the rotation lines
in the band are equally spaced and can be reordered in frequency so
that they form an array with exponentially decreasing line intensities
from the band center. For this reason, the model is called the exponential
wide band model. Tien and Lowder (ref. 28) have presented band correla-
tions with a single continuous correlation equation for all mass path
lengths. The correlation is based on the construction of a function that
meets all the mathematical requirements of the absorptance as a function
of mass path length.
The exponential wide band model has been used to obtain corre-
lation constants, and these have been gathered together by Edwards


broadening Lower limit Upper limit Effective band width,
parameter, of AT of d AT
C?P, ?> ?>
p=-- CIII-l cm-'

4C I C:,

0 PC3 A= C,S

PC, c:1(2-P) A= C , ( X P , )'Iz-pC:j

p s 1
Cz(2-PI co 4G.i
A=Cn(ln --+Z -P)

0 C:3 A= C,S
Cn n

" C , , C,. C,. 6, and rr are ill tablr 5-111. S i a mass pall, length, pS, g/rn2. P,= [(F+ bpy,)/P,]" where P,=1 atm, 11 is
ljartial pressure of absor1,ing gas, and pv, is partial pressure of !VL broadening gas i n atmosl~hercs.

e t al. (ref. 29) from the references listed in table 4-1. The resulting effec-
tive band widths and upper and lower limits of the A, can be obtained
from the relations in table 4-11. These results are given in units of wave
number which is cm-'. The quantities b, n, C1, Ce, and C:$ needed to
evaluate these relations are given in table 4-111 for COe, CH4, HnO, and
CO in a mixture with nitrogen. The method of using these band correla-
tions will be shown by two example problems.

Gas pm

,I 41.2
3.0poll (7')
2.5plol ( T )
,55( TIT,,)".'

65 ( 7'/7;,)"."
46 (TIT,,)"-"

"For lirnits on 7' and X,see table 4-1

'"j-,, 1s ,
aken as I00 K for all r a s r s .
For C O 1 ,

"For HrO,

' For CO,

q.= [l.5.l5+0.22 (E)'"] [ I - ? ~ ~ ( - ~ ) ]

EXAMPLE4-1: Find the effective band width 2 of the 9.4-pm band of

pure COz at 1 atm and 500 K for a path length S of 0.364 m.
To obtain 2 from the relations in table 4-11, the constant C1 must be
evaluated. From table 4-111 at the 9.4-pm COz band


Substitute the values vl = 1351 cm-', 77, = 667 cm-l, 773 = 2396 em-',
h=6.625X (erg) (sec), k z 1 . 3 8 0 X 10-l6 erg/K, c=2.998X 10''
cmlsec, and T=500 K. This gives (pl=0.0196 so that C1=0.0149
m"(cm)(g). Table 4-11 gives the quantity /3 as /3=CgP,/4CICz. For the
9.4-pm COz band, table 4-111 gives the Cz and C3 as


so that

From table 4-11, P , for pure COP at 1 atm is

Then, since To = 100 K


Because p s 1, the correlation equations for the specified conditions

are the first set in table 4-11. The mass path length is given by

The gas density is

liter ) (%)=
g-mole ( looO
P= 103 -
liter m3
22.42 ---

so that the mass path length is

The choice of correlation equation depends on the limits into which X

causes A to fall. The first equation in table 4-11 gives

but this falls well outside the prescribed upper limit of the band given by

for the p < 1 part of the correlation. For intermediate X, the second
line of table 4-11 gives

A= C2 (XP,) -PC3

and this lies within the range

for this part of the correlation. The result for A compares reasonably well
with an experimental value of 5.9 cm-' from reference 20 for similar

EXAMPLE4-2: Determine the energy per unit area and solid angle
of the 9.4-,urn band emitted from the end of a thin column of CO2 gas
at 1 atm pressure and 500 K if the column is 0.364 m long.
Using equation (4-6213) and integrating only over the 9.4-t~mband
result in

where equation (4-73) has been substituted and n b n ~ t d c e , l f e ris the wave
number of the band center. For this band table 4-111 gives

Using equation (2-llc) of volume I for iLb a n d l from example 4-1 gives
the result

The preceding discussion has been for single component gases. If

two gases are present and both absorb energy, then the b a n d absorp-
tance of each may overlap in some spectral regions. In this c a s e , Hottel
and Sarofim (ref. 30) show that, for two gases, a and b, in an overlapping
band of width A n , the following relation is valid:

Thus the simple sum of the two 2 is reduced by the quantity A @ A r ) .

Restriction is to wave number intervals over which both A a d A b are
applicable average values, and in which there is no correlatiom between
the positions of the individual lines of gases a and 6.
Many additional complexities are introduced when a gas m i x t u r e is

considered. For example, the partial pressure p of absorbing gas in a

multicomponent system varies with T and P, the populations of the
energy states vary with T, and the overlapping of spectral lines changes
with P. It is thus very conlplex to analytically fornlulate the dependence
of on T, p, and P for a real gas mixture. Useful results must depend
heavily on experiment while using theory as a guide.
Hottel and Sarofim (ref. 30) discuss in detail total absorptance curves
of the type shown in figure 1-11. Such curves are available for a number
of gases, and their accuracy has been confirmed by many recent meas-
urements. The use of total absorptances and effective band widths for
various engineering problems will be discussed in chapter 5.


In this chapter, some consideration was given to microscopic absorp-
tion phenomena in gases. Derivations based on the microscopic ideas
were related to some concepts already developed on the macroscopic
basis-Planck's spectral distribution, induced emission, the equation
of transfer, and the absorption coefficient. In addition, some framework
was constructed concerning the radiative band absorption properties
of polyatomic gases and their dependence on path length and density.
The material on gas properties will be used in chapter 5 in conjunction
with some engineering approaches for calculating radiative transfer
in common gases.

1. EINSTEIN, ALBERT: On the Quanta Theory of Radiation. Phys. Zeits., vol. 18, Mar. 15,
1917, pp. 121-128.
2. EINSTEIN, ALBERT: Emission and Absorption of Radiation According to the Quantum
Theory. Verh. deut. phys. Ges., vol. 18, 1916, pp. 318-323.
3. HEITLER, WALTER:The Quantum Theory of Radiation. Third ed., Clarendon Press,
Oxford, 1954, pp. 412-414.
of Gas Dynamics. Addison-Wesley Pnbl. Co., 1965.
5. WIESE, W. L.; SMITH, M. W.; AND GLENNON, B. M.: Atomic Transition Probabilities.
Vol. I: Hydrogen Through Neon-A Critical Data Compilation. Rep. NSRDS-NBS-
4, Vol. 1, National Bureau of Standards, May 20, 1966.
6. GRIEM, HANSR.: Plasma Spectroscopy. McGraw-Hill Book Co., Inc., 1964.
7. BREENE,ROBERTG.: The Shift and Shape of Spectral Lines. Pergamon Press, 1961.
8. PENNER, S. S.: Quantitative Molecular Spectroscopy and Gas Emissivities. Addison-
Wesley Publ. Co., 1959.
9. ARMSTRONG, B. H.; ET AL.:Radiative Properties of High Temperature Air. J. Quant.
Spectrosc. Radiat. Transfer, vol. 1, no. 2, 1961, pp. 143-162.
10. GOODY, R. M.: Atmospheric Radiation. Vol. 1. Theoretical Basis. Clarendon Press,
Oxford, 1964.

11. EI.SASSER,WALTERM.: Heat Transfer by Infrared Radiation in the Atmosphere. Har-

vard Meteorological Studies No. 6, Harvard University, 1942.
12. PLASS, GILBERTN.: Useful Representations for Measurements of Spectral Band
Absorption. J. Opt. Sac. Am., vol. 50, no. 9, Sept. 1960, pp. 868-875.
13. TIEN,C. L.: Thermal Radiation Properties of Gases. Advances in Heat Transfer. Val. 5.
T. F. Irvinr, Jr., and J . P. Hartnett, eds., Academic Press, 1968, pp. 253-324.
14. EDWARDS, D. K.; A N D MENARD,W. A,: Comparison of Models for Correlation of Total
Band Absorption. Appl. Opt., vol. 3, no. 5, May 1964, pp. 621-625.
15. KYLE,T. G.: Absorption of Radiation by Uniformly Spaced Doppler Lines. Astrophys.
J., vol. 148, no. 3, June 1967, pp. 84.5-848.
16. GOLDEN,S. A.: The Doppler Analog of the Elsasser Band Model. J. Quant. Spectrosc.
Radiat. Transfer, vol. 7, no. 3, 1967, pp. 483-494.
17. GOLDEN,S. A,: The Doppler Analog of the Elsasser Band Model. 11. J. Quant. Spec-
trosc. Radiat. Transfer, vol. 8, Mar. 1968, pp. 877-897.
18. GOLDEN,S. A.: The Voigt Analog of an Elsasser Band. J. Quant. Spectrosc. Radiat.
Transfer, vol. 9, no. 8, Aug. 1969, pp. 1067-1081.
19. EDWARDS,D. K.: Radiant Interchange in a Nongray Enclosure Containing an Iso-
thermal Carbon-Dioxide-Nitrogen Gas Mixture. J. Heat Transfer, vol. 84, no. 1,
Feb. 1962, pp. 1-11.
20. EDWARDS,D. K.: Absorption of Infrared Bands of Carbon Dioxide Gas at Elevated
Pressures and Temperatures. J. Opt. Soc. Am., vol. 50, no. 6, June 1960, pp. 617-626.
21. EDWARDS,D. K.; AND MENARD,W. A.: Correlations for Absorption by Methane and
Carbon Dioxide Gases. Appl. Opt., vol. 3, no. 7, July 1964, pp. %7-852.
22. EDWARDS, D. K.; AND SUN,W.: Correlations for Absorption by the 9.4-p and 10.4-p
Con Bands. Appl. Opt., vol. 3, no. 12, Dec. 1964, pp. 1501-1502.
23. EDWARDS, D. K.; FLORNES,B. J.; GLASSEN,L. K.; AND SUN,W.: Correlation of Absorp-
tion by Water Vapor at Temperatures from 300°K to llOO°K. Appl. Opt., vol. 4, no. 6,
June 1965, pp. 715-721.
24. EDWARDS,D. K.: Absorption of Radiation by Carbon Monoxide Gas According to the
Exponential Wide-Band Model. Appl. Opt., vol. 4, no. 10, Oct. 1965, pp. 1352-1353.
25. EDWARDS,D. K.; AND NELSON,K. E.: Rapid Calculation of Radiant Energy Transfer
Between Nongray Walls and Isothermal H 2 0 or C 0 2 Gas. J. Heat Transfer, vol. 84,
no. 4, Nov. 1962, pp. 273-278.
26. WIENER,MICHAELM.: Radiant Heat Transfer in Nnn-isothermal Gases. Ph.D. Thesis,
Univ. California at Los Angcles. 1966.
27. HINES, W. S.; AND EDWARDS,D. K.: Infrared Absorptivities of Mixtures of Carbon
Dioxide and Water Vapor. Chem. Eng. Progr. Symp. Ser., vol. 64, no. 82,1968, pp.
28. TIEN, C. L.; AND LOWDER,J. E.: A Correlation for Total Band Absorptance of Radiat-
ing Gases. Int. J. Heat Mass Transfer, vol. 9, no. 7, .July 1966, pp. 698-701.
Heat Transfer in Nonisothermal Nongray Gases. J. Heat Transfer, vol. 89, no. 3,
Aug. 1967, pp. 219-229.
30. HOTTEL,HOYTC.; AND SAROFIM, ADELF.:Radiative Transfer. McGraw-Hill Book Co.,
Inc., 1967.
mission Through Synthetic Atmospheres. Geopbys. Res. Papers No. a, AFCRL-
TR-55-213, Air Force Cambridge Research Center, 1955.

32. HOWARD,J . N.; BURCH,D. E.; A N D WILLIAMS,DUDLEY:Infrared Transmission of

Synthetic Atmospheres. I. Instrumentation. J. Opt. Sac. Am., vol. 46, no. 3, Mar.
1956, pp. 186-190.
33. HOWARD,J. N.; BURCH,D. E.; AND WILLIAMS,DUDLEY:Infrared Transmission of Syn-
thetic Atmospheres. 11. Absorption by Carbon Dioxide. J. Opt. Sac. Am., vol. 46,
no. 4, Apr. 1956, pp. 237-241.
34. HOWARD.J. N.; BURCEI,D. E.; AND WILLIAMS,DUDLEY:Infrared Transmission of
Synthetic Atmospheres. IV. Application of Theoretical Band Models. J. Opt. Sac.
Am., vol. 46, no. 5, May 1956, pp. 334-338.
35. STULL,V. ROBERT;AND PLASS,GILBERTN.: Spectral Emissivity of Hydrogen Chloride
from 1000-3400 cm-'. J. Opt. Soc. Am., vol. 50, no. 12, Dec. 1960, pp. 1279-1285.
36. AROESTE,HENRY;AND BENTON,WILLIAMC.: Emissivity of Hydrogen Atoms at High
Temperatures. J. Appl. Phys., vol. 27, no. 2 , Feb. 1956, pp. 117-121.
Chapter 5. The Engineering Treatment of Gas Radiation
in Enclosures

A11 extensive body of engineering literature exists dealing with radia-

tion exchange between solid surfaces when no absorbing medium is
present between them. The methods for treating such problen~sare
highly developed, and have been examined at length in volumes I and
I1 of this publication (refs. 1 and 2). The additional complication of
having an intervening absorbing-emitting gas present in problems of
energy exchange between surfaces can be accounted for by building
upon the foundation established for the simpler problems. In this
chapter, the relations developed in chapters 1 to 4 are used in the
derivation of engineering methods for solving gas-radiation problems.
These methods are a direct extension of the surface-surface energy
exchange methods developed for enclosures in volume 11. The engineer
familiar with surface exchange'analyses will then find that much of the
nomenclature, and the physical intuition that he possesses, can be carried
over to gas radiation problems.
Most of the material in this chapter will be concerned with an
absorbing-emitting gas that is isothermal. As compared with the develop-
ment in chapter 3, this provides the simplification that the gas tem-
perature distribution need not be computed to obtain the radiative
behavior of the gas. In section 5.8 some of the methods developed for the
isothermal gas will be carried over to nonisothermal gas computations.

A area
AFG geometrical absorption factor
AF? geometrical transmission factor
2 effective band width
a absorption coefficient
a, b, c dimensions in system of two rectangles
C ratio Le/Le,,

pressure correction coefficients

band coefficient in equation (5-76)

D spacing between parallel plates; diameter
E,Y (1- E N ) /E,V

exponential integral
emissive power
geometric configuration factor
exchange factor
gas-gas direct exchange area
gas-surface direct exchange area
height of cylinder
radiation intensity
mean beam length of gas volume
mean beam length for limiting case of small absorption
total number of surfaces in enclosure
total pressure of gas or gas mixture
partial pressure
energy per unit time
energy flux; energy per unit area and time
radius of hemisphere, semicylinder, cylinder, or sphere
coordinate along path of radiation
geometric mean beam length
surface-gas direct exchange area
surface-surface direct exchange area
absolute temperature
width of plate
mass path length, pS; shortest dimension of rectangular
a( S ) absorptance
5 (S) geometric mean absorptance
A a , A€ correction for spectral overlap
P cone angle, angle from normal of area
akj Kronecker delta
E emissivity of surface
E(S) emittance of medium
"I wave number
K optical depth
h wavelength
P reflectivity; density
0- Stefan-Boltzmann constant
T(S) transmittance
+(S) geometric mean transmittance
W solid angle

b blackbody

C 0, carbon dioxide
g gas
Hz0 water vapor
i incident, incoming
j, k surfaces j or k
j-k from surface j to surface k
1 absorption band 1
o outgoing
u uniform
w wall
A spectrally (wavelength) dependent
rl wave number dependent
directional quantity
+ quantities defined after equation (5-67)
* **7 dummy variable of integration


In section 5.3 of volume I1 the radiation exchange equations were
developed for an enclosure that did not contain an absorbing-emitting
medium, and that had surfaces with spectrally dependent properties.
Since the absorption properties of gases and other absorbing media
are almost always strongly wavelength dependent, the present develop-
ment will be carried out at a single wavelength. Then in a later section
integrations will be performed over all wavelengths to obtain the total
radiative behavior. As in most of the development of volume I1 it will
be assumed that surface directional property effects are sufficiently
unimportant that the surfaces can be treated as diffuse emitters and
Often in a gas filIed enclosure such as in an engine combustion
chamber or industrial furnace, there is sufficient mixing so that the
entire gas is essentially isothermal. In this instance the analysis is
simplified by the fact that it is unnecessary to compute the gas tempera-
ture distribution. Eve11 with this sin~plification,however, a detailed
radiation exchange computation between the gas arid bounding surfaces
is quite involved.
Consider an enclosure composed of N surfaces, each at a uniform
temperature as shown in figure 5-1. Typical surfaces are designated
by k and j. The enclosure is filled with an absorbing emitting medium
at uniform temperature T,. The quantity Q, is the amount of heat that

5-1.-Enclosure cornposed of N discrete surface areas and filled with uniform
gas g (enclosure shown in cross section for simplicity).

it is necessary to supply by means other than radiation to the entire

absorbing medium in order to maintain this temperature. A common
source of the Q, would be by combustion. If in the solutioil of a problem,
Q, comes out to be a negative number, the medium is gaining a net
amount of radiative energy from the enclosure walls and the energy
must be removed from the gas to maintain it at its steady temperature
T,. The Q, is analogous to the Q/< in volume I1 which is the energy
supplied by some external means to area A k .
The enclosure theory will yield equations relating the Qk and Tk for
each surface and the Q, and T, of the gas or other absorbing isothermal
medium filling the enclosure. Considering all the surfaces and the gas, if
half of the Q's and T's are specified, then the radiative heat balance
equations can be solved for the remaining unknown Q or T values. If the
heat input to the gas from external sources Q, is given, the analysis will
yield the steady gas temperature T,,. Conversely, if Tf, is given, the anal-
ysis will yield the energy that must be supplied to maintain this gas
The net radiation method as developed in chapters 3 and 5 of volume
I1 will now be extended to include gas radiation terms. At the kth surface
of an enclosure as shown in figure 5-2 a heat balance gives

FIGURE5-2.-Spectral energy c~uantitiesincident upon and leaving typical surface area of


The dqx,, and d q ~ ir,, are respectively the outgoing and the incoming
energy fluxes in a wavelength interval dA. The c ~ Q A k ,is the energy

supplied to the surface in the wavelength region dh. Note that as dis-
cussed in connection with equation (5-4) of volume 11, the external
energy supplied to A,,. is

1:;. d ~ A %

The outgoing spectral flux is composed of emitted and reflected energy

The functional notation will usually be omitted in order to shorten the

form of the equations that follow. The eAb, kdX is the blackbody spectral
emission at Tk in the wavelength region d h about the wavelength A.
The dqhi, k in equation (5-1) is the incoming spectral flux to Ak. It is
equal to the sum of the contributions from all the surfaces that reach
the kt" surface after allowing for absorption by passing through the
intervening gas, plus the contribution by emission from the gas. The
equation of transfer allows for both attenuation and emission as radiation
passes along a path through the gas. A typical path from Aj to Ak within
an incident solid angle dok is shown in figure 5-1. If all such paths and
solid angles are accounted for by which radiation can pass from all the
surfaces (including Ak if it is concave) to Ak, the solid angles dwk will
encompass all of the gas region that can radiate to Ak. Thus, if the
equation of transfer which includes the gas emission term is used to
compute the energy transported along all paths between surfaces, the
gas emis'sion will automatically be included. The radiation passing from

one surface to another, including emission and absorption by the inter-

vening gas, will now be considered.
A typical pair of surfaces is shown in figure 5-3. In the enclosure
theory dqho is assumed uniform over each surface. Since the surfaces
are assumed here to be diffuse, the spectral intensity leaving dAj is
ii,,j= d q ~ ~ , j / ( ~ dBy
X )use
. of the equation of transfer (eq. (2-lo)), the
intensity arriving at dAk after traversing the path S is

iLi,j-k=i;o,j eXP (-KA) + /:Ai;b,,(K:) exp [- (KA-K:)]~K: (5-3)


is the optical depth along the path S . The gas is assumed to be at uniform
temperature and to have a constant spectral absorption coefficient.
Equation (5-3) then reduces to

i;i,j.k = i;,, exp (- U A S + ,

, exP [-
) a ~ i ; ~JOs (I,A ( S - S *) ] W e

which can be further integrated to give

iLi,j.k= il;,, exp (- U A S ) + - exp (- I (5-4)

Now for convenience introduce the definitions that T A ( S )= exp (- axS)

is the spectral transmittance of the gas of path length S , and a x ( S )
1 - exp (- AS) is the spectral absorptance along the path. Then equation
(5-4) is written as

This intensity arriving at dAk in the solid angle dwk provides an arriving
energy equal to i);i,j-kdAkcos PkdwkdX But dwk=dAj cos Pj/S2 SO that
the arriving spectral energy is

d 3 Q ~j-li=
j , iLi,j-kdAkdAj dX
dAkdAj cos PI; cos pj
= rib,jrA(S) + i;b,{g(TY)a~(S)l S2
dX (5-6)

For a diffuse surface dqx,, j= ~ i ; , jdX,

, and also e ~ bY ,= Ti);*,y. SO that

5-3.-Radiation between two surfaces with isothermal gas between them.

equation (5-6) can be written as

Equation (5-7) is now integrated over all of AI, and Aj to give the
spectral energy along all paths from Aj that is incident upon An,

The first term of the double integral is the spectral energy leaving Aj
that is transmitted to Ah.. The second term is the spectral energy received
at AI, as a result of emission by the constant temperature gas in the
envelope between Aj and An. This envelope is the volume occupied
by all straight paths between any part of Aj and An.

5.3.1 Definitions of Spectral Transmission and Absorption Factors

The double integration in equation (5-8) has some similarity to the
double integral in equation (2-22) of volume I1 for the configuration
factor between two surfaces without an intervening gas. By analogy
define the factor 7 ~ ~ j - 1such
, . that

where Fj-k is the geometric configuration factor with no absorbing

medium as used in volume 11. With no absorbing medium present,
TA(S)= 1 and the right side of equation (5-9) becomes F j - k . Hence, in
this instance yj-k=l. The YA, j-~,.is called the geometric mean trans-
mittance from Aj to Ak. Similarly from the second quantity in the bracket
of equation (5-8), a geometric mean absorptance % A , j-k is defined as

For a nonabsorbing medium (Y~.j-k=o,while for perfect absorption

a ~ , ~ - k 1.
= From the definitions of T A and a x , and equations ( 5 9 ) and
( 5 1 0 ) ,the 7~ and are related by

An alternate terminology is also used wherein the entire quantity

AjFj-k7h,j-k is called the geometrical transmission factor, and the
quantity AjFj-k(YA,j-k is called the geometrical absorption factor.
Equation (5-8) can now be written as

When computing the heat exchange in an enclosure, it will be neces-

sary to determine the ?A and 6 x . This usually involves some difficult
double integrations. In the present discussion the enclosure theory formu-
lation will first be completed. Then the evaluation of ?Aand alAwill be con-
sidered. It is only necessary to perform one double integration to obtain
bcith the ?Aand (YA because of the relation in equation (5-11).

5.3.2 Matrix of Enclosure Theory Equations

For an enclosure with surfaces bounding an isothermal gas at To,
the incident spectral energy on any surface Ak will equal that arriving
from the directions of all surrounding surfaces, which gives

From reciprocity (eq. (2-25) of vol. 11) AjFj-k=AkFk-j SO that the Ak can
be eliminated to give

Equatioils (5-l), (5-2), and (5-14) form a set of three equations in

three unknowils dqAo, dqAi, and dqh for each of the surfaces in the
enclosure. The dqhi is eliminated by combining equations (5-1) and
(5-2) and also by substituting equation (5-14) into equation (5-1). This
yields the set of two equations for each surface

Equation (5-15) is the same as for an enclosure without an absorbing

gas (see eq. (5-8) of vol. I1 for example). Equations (5-15) and (5-16)
are analogous to equations (3-6) and (3-7) of volume I1 for the simpler
case of a gray enclosure without an absorbing gas. From the symmetry
of the integrals in equations (5-9) and (5-10) and the reciprocity relation
AjFj-I;= AkFk-j, it is found that
?A, j-x.= 7 h . k-j (5-17)

ah, j-k= ( Y A , k-j (5-18)

Then equation (5-16) can also be written as

As in section of volume 11, the set of equations (5-15) and

(5-19) can be further reduced by solving equation (5-15) for dqh, and

inserting it into equation (5-19). This results in the relation

The Kronecker delta 6kj has the values: 6kj = 1 when /i=j, and 8 k j = 0
when k # j. This equation is analogous to equation (3-19) of volume 11.
If equation (5-20) is written for each k from 1 to N, a set of N equations
is obtained relating the 2N quantities d q ~and e A b for the surfaces,
since the gas temperature and hence ebb,, is assumed known. One-half
of the dqh and exb values have to be specified and the equations can
then be solved for the remaining unknowns. To determine the total
energy quantities, this set of equations would have to be solved in a
number of wavelength intervals and the integration of each quantity
performed over wavelength.

5.3.3 Heat Balance on Gas

Before discussing the solution of the enclosure equations in more
detail, there is an additional heat balance that is of interest. This is a
heat balance on the gas, which will provide the energy required to main-
tain the gas at the specified temperature. From an energy balance on the
entire enclosure, the energy that must be supplied to the gas by com-
bustion, for example, is equal to the net quantity escaping from the
boundaries. The total energy escaping from the enclosure at surface k is

Then the energy added to the gas is found by summing over all surfaces;
that is,

This can be evaluated after the dqh are found for each surface in a suf-
ficient number of wavelength intervals from the matrix of equation

EXAMPLE5-1: As an example of the net radiation method consider

the heat transfer in a system of two infinite parallel plates at temperatures
TI and TI?(TI > T;L) separated by a gas at uniform temperature Tg.
Equation (5-20) applied to a two surface enclosure gives (note that
F1-l=FP-2=O) for k = 1 :

-F1- 2 7 ~ .l - z e ~ b ,p d h -F gdA
1 - 2 6 ~ , - ~ e ~ b , (5-22a)
and for k = 2:

For the infinite parallel plate geometry, F1-2=F2-1=1,and from

equations (5-17) and (5-18) ?A, 2 - 1 = ?A, and 6 ~2 - t, = 6 ~l - 2,. For
simplicity the numerical subscripts on the ? and 6 will be omitted. Then
equations (5-22a) arid (5-22b) become

1 I - E A , ~ -
-d q h , 1 ----r ~ d q z ~=, ( e ~ b I, - ?Aehb, 2 -G g)dA
A ~ A ~ , (5-23a)
€A, t EX, 2

Equations (5-23a) and (5-2323) are solved simultaneously for d q ~1, and
After rearrangement and using the relation ( E A = 1- ? A , this yields
d q ~ 2.,

The total energy fluxes added to surfaces 1 and 2 are, respectively,

41 = rrn
A = O&A, 1 and q, = jzA=O

The total energy added to the gas in order to maintain its specified
temperature is equal to the net energy removed from the plates. Hence,
per unit area of the parallel plates

When the medium between the plates does not absorb or emit radiation,

then ?x= 1 and ec~uations(5-24a) and (5-2413) reduce to equation (5-10)

of volume 11. With an absorbing-radiating gas present the numerical
integration of equations (5-24a) and (5-24b) over all wavelengths to
obtain the ql and 9.2 is difficult because of the very irregular variations
of the gas absorptiol: coefficient with wavelength. The integration will
be further discussed in section 5.7 by dividing the wavelength range
into bands of finite width that are either absorbing or nonabsorbing.


To compute values from the exclzailge equations, the quantities
? and 5 or (AFT) and (AFc) must be evaluated. By use of the definitioils
given in equations (5-9) and (5-10)

~ j ~ j - k j ~- k i= l, q k llj
[ I - exp (- clxs); cos P k

It is evident that it is the double integral in equation (5-27) that must be

carried out for various geometrical orientations of the surfaces Aj and
At. The evaluation for some specific geometries will now be considered.

5.4.1 Hemisphere to Differential Area at Center of I t s Base

As shown in figure 5-4, let Aj be the surface of a henlisphere of radius
R , and dAk be a differential area at the center of the hemisphere base.

FIGURE5-4.- Hemisphere filled with isothermal gas.


Then equation (5-27) becomes, since S = R and ,Bj=O (path R is normal

to hemisphere surface),

The convenient dAj is a ring element dAj=2rrR2 sin PkdPh., and the
factors involving R can be taken out of the integral as R is constant for
the hemisphere geometry. This ,j ves

= dAk exp (- ahR)

By using AjdFj-dk=dAkFdk-j and noting that Fdk-j= 1, this reduces to

?h,j-d,,.= exp (- O A R ) (5-29)

This especially simple relation will be used later in conjunction with the
concept of mean beam length. This is an approximate technique wherein
the radiation from an actual gas volume is replaced by that from an
effective hemisphere of gas.

5.4.2 Top of Right Circular Cylinder to Center of Its Base

This geometry is shown in figure 5-5. Since Pj= P k = P , the integral
in equation (5-27) becomes for the top of the cylinder Aj radiating to the
element at the center of its base dAk

A convenient change in the integral is made by noting that dAj cos PIS2
is the solid angle by which the ring dAj is viewed from dAk. By considering
the intersection of this solid angle with the surface of a unit hemisphere,
it is found that this solid angle is also equal to 2rr sin P dp. Making this
substitution results in the integral in equation (5-30) being transformed
so that
AjdFj--dE;K,j-d,,.= dAk
I, [exp ( - a ~ S ) ] 2cos P sin P dp (5-31)

Now let ( I A S = K AThen. from figure >5 cos P= ~ / S = ~ C L A /and KA,

sin p dp=- d(cos p ) = ( k a ~ / ~ f As ) d P~goes
~ . from 0 to p,,,the limits on
the variable K A are K A = N A / I to ( l A Then equation (5-31) becomes

5-5. -Geometry for exchange from top of gas-filled cylinder to center of its base.

This can be integrated by parts two times to yield

A j d F j - d k ? ~ ,j-dk= (aih) ' d A k ( K,,

a , i . h \ m

n ~ h

The last integral is a tabulated exponential integral function so the

result can be evaluated without difficulty for various values of the
parameters Rlh and a k h .
The integral in equation (5-32) can also be found directly in terms of
the exponential integral function defined in equation (2-45) by writing

By letting K A = ( C ~ A - ) / ~ and axhlp, respectively, in the two

integrals, they become

The integral in equation (5-32) can then be written in terms of the

exponential integral function as

5.4.3 Side of Cylinder to Center of Its Base

Let dAj be a ring around the wall of a cylinder as shown in figure
5-6, and note that dAj cos pj/S2 is the solid angle by which dAj is viewed

5-6. -Geometry for exchange from side of gas-filled cylinder to center of its base.

from dAk. This solid angle is also equal to 27-r sin P s dP,,. Then equatioll
(5-27) can be written for the side of the cylinder to dAk as

= 2dAk
I, [exp ( - a ~ s ) ]cos Pk sin PI,. (@I; 6-35)

This is of the same form as equation (5-31). Let a k S = K A ; then sin P k =

R / S = R U A / K and
A cos f i k dPk = d(sin P k ) =- ( R c ~ A / )K~: K A Making
. these
substitutions and integrating by parts as for equation (5-33) yield

nih d(RIh)'+l (5-36a)

nil8 ( H l h )

Alternatively by use of equation (5-34)

A s for equation (5-33) (or (5-34)), this result can be readily evaluated for
various values of the parameters Rllz and ahh.

5.4.4 Entire Sphere to Any Element on Its Surface or to Its Entire Surface
From figure 5-7 since pk = pj let them both be simply P. Then S = 2 R
cos p , and using the form in equation (5-31) gives

A ~ ~ F ~ - , ~ ; A ,
exp 2R
( - ~ A s ) s ~ s


5-7.- Geometry for exchange of surface of gas-filled sphere to itself.

Integrating gives

AjdFj-dk'Ti,j-dk=---- 2d4k [ I - (2aiR
+ 1 ) exp (-ZCLAR)] (5-37)
which is in terms of only the single parameter 2aiR.
Equation (5-37) can be integrated over any finite area A t to give the
7~ from the entire sphere to Ak as

- 2Ak
AjFj-k~i,j-k =
[1 - ( 2 a ~ R 1 ) exp ( - ~ u A R ) ]

Since Fj-k=Ah./Aj (from eq. (3-65) of vol. I I ) ,

- 2
' T i , j-I<= [l- (2nhR+ 1 ) exp ( - 2 a i R ) ] (5-38)

which also holds for the entire sphere to its entire surface.

5.4.5 Infinite Plate to Area on Parallel Plate

If there is considered on one plate an element dAk (fig. 5-8) and on
the other plate a concentric ring element dAj centered about the normal
to dAk, the geometry is like that in figure 5-5 for a ring on the top of a
cylinder to the center of its base. Then from equation (5-32)

5-8. -1sother1nal gas layer between infinite parallel plates.

where D is the spacing between the plates. By use of the procedure

leading to equation (5-34), the integral is transformed to E3(ahD)/(ahD)2.
Then integrating over any finite area Ax. as shown in figure 5-8 gives

By using AjFj-k=AkFk-j and noting that Fk-j= 1, this reduces to

5.4.6 Rectangle to a Directly Opposed Parallel Rectangle

Consider a s in figure 5-9 the exchange from a rectangle to a n area
element on a directly opposed parallel rectangle. The upper rectangle
has been divided into a circular region and a series of partiaI rings of
small width. The contribution from the circle of radius R to
AjdFj-dli?~,j - d ~ . can be found from equation (5-33) which is for t h e top
of a cylinder to the center of its base. For the nth partial ring, let fll be
the fraction it occupies of a full circular ring. Then by use of equation
(5-31), the contribution of all the partial rings to AjdFj-dk'^;h,j-cl~is
approximated by

dAk fll exp (-ahS1,) 2 cos PI, sin PI, A/31,


This evaluation of AjdFj-dkSh,j-dk is carried out for several area


/ ' nth ring

. ' '\ / 'Aj

5-9.-Geometry for exchange between two directly opposed parallel rectangles
with intervening gas.

patches on Ak. This is usually sufficient so that the integration over Ak

can be performed as indicated by equation (5-9) to yield

In some practical situations it is desired to determine the radiant
energy from a mass of isothermal gas to all or part of its boundaries,
without considering emission and reflection from the boundaries. An
example would be radiation from hot furnace gases to walls that are cool
s o that emission is small, and that are rough and contaminated with soot
s o that they are essentially nonreflecting. In equation (5-13) the dqt,o,j,
which is the spectral outgoing heat flux from a typical surface A j , would
then be zero. The spectral incoming energy at surface Ah. is then

If the geometry consists of a hemisphere of gas radiating to an area

element dAk at the center of its base as shown in figure 5-4, equation
(5-40) has an especially simple form. Since the hemispherical boundary
is the only surface in view of dAk, and dAk is a differential element,
equation (5-40) reduces to

From equation (5-29),

o ( ~ , ~1 -- 7~~j ~dk=
-~ = 1- exp ( - a d )

Note also that for radiation between the surface of a hemisphere and the
center of its base, Fdk-j= 1 so that froill reciprocity dFj-dk=dAk/Aj. By
combining these results, equation (5-41) reduces to the following sirllple
expression giving the incident heat flux from a hemisphere of gas to
the center of the hemisphere base,

& x i , k= [I - exp ( - a h R ) ] e ~ h,dA

, (5-42)

From equation (5-42), 1- exp (- aAR) is the spectral emittance of the

gas E X ( A, T , P , R ) for path length R.7 Then equation (5-42) becomes

Thus a very simple forin is obtained for the energy incident upon clAk
from the hemisphere of gas of radius R surrounding c/Ak. The incident
energy depends on the optical radius of the hemisphere axR.

It would be most convenient if a relation having the simple form of

equation (5-43) could be used to determine the value of dq~i,kon A t
for any geometry of gas volume radiating to all or part of its boundary,
rather than only for a hemisphere radiating to the center of its base.
Because the geometry of the gas enters equation (5-43) only through
E A ( C L AitR )is, possible to define a fictitious value of R , say L,, that would
give a value of ~x(axL,)such that equation (5-43) would give the correct
dqii for another geometry. This fictitious le~lgthL, is called the mean
beam length. Then for an arbitrary geometry of gas let

The mean beam length is thus the required radius of a gas hemisphere
such that it radiates a flux to the center of its base equal to the average
flux radiated to the area of interest by the actual volunle of gas.

'For simplicity the prime notation used for a directional quantity will be omitted; in this instance the E L is indrgendent of

5.5.1 Mean Beam Length for Gas Between Parallel Plates

Radiating to Area on Plate
Consider for example the geometry of two black infinite parallel plates
at zero absolute temperature separated by a distance D. The plates en-
close a uniform gas at temperature T,, with absorption coefficient ([A.
The rate at which spectral energy is incident upon A k on one plate
(fig. 5-8) is from equations (5-40) and (5-39)

Since the plates are infinite, F k - j = 1 . Then by reciprocity (eq. (2-25)

of vol. 11), F j - k = A k / A j and equation (5-45) reduces to

Comparing equations (5-46) and (5-44) reveals the mean beam length to

o r in terms of the optical thickness UAD

5.5.2 Mean Beam Length for Sphere of Gas Radiating to Any Area on
Consider gas in a nonreflecting sphere of radius R where the sphere
boundary Aj is at Tj= 0. By use of equations (5-40) and (5-37) the radia-
tion incident on an element dA,; is

X + 1
[ l - ( 2 a ~ R 1 ) exp (- 2 a ~ R ) ]

For a sphere dFj-dk= dAk/Aj(by use of eq. (3-64) of vol. 11). Then

Equate this to d q from

~ equation (5-44) to obtain
X [ l - (2ahR + 1 ) exp (-2a~R)]
1- exp (- aAL,) = 1-
[ l - ( 2 a ~ R 1 ) exp (- 2 a ~ R ) ]

which gives

-- I
{ (2aAR)' [ l -(2a~R + 1 ) exp ( - 2 a ~ R ) ]I (5-48)

In view of the general applicability of equation (5-38), equation (5-48)

gives the correct mean beam length for the entire sphere radiating to
any portion of its boundary.

5.5.3 Radiation from Entire Gas Volume to Its Entire Boundary in Limit
When Gas I s Optically Thin
Because of the integrations involved, the mean beam length will
usually be difficult to evaluate. It is fortunate that some practical approxi-
mations for the mean beam length can be found quite simply by looking
at the limit when the gas is optically thin. In the optically thin limit, by
expanding the exponential term in a series for small aAS, the transmit-
tance ~ ~ = e x(-pAS) becomes

( aAS)'
ah+ o
TA= lim
arS+ o
. . .]=I

Any differential volume of the uniform temperature gas emits the spec-
tral energy 4 ~ h e h b , ~ d h d VSince
. r ~ =
1, there is no attenuation of the
emitted radiation and all of it reaches the enclosure boundary. For the
entire radiating volume the energy reaching the boundary is 4 a ~ e ~gdAY b,
so that the average spectral flux received at the boundary of entire area
A is

By use of the mean beam length the average flux reaching the boundary
is from equation (5-44),

For the special case of emall absorption let L , be designated b y L,,,.

Then expand the exponential term in equation (5-50) in a series to obtain
for small UAL,,,

Equating this to the dqii in equation (5-49) gives the desired result for
the mean beam length of an optically thin gas radiating to its entire

To give a few examples, for a sphere of diameter D

For an infinitely long cylinder of diameter D

For gas between infinite parallel plates spaced D apart

5.5.4 Correction for Mean Beam Length When Gas Is Not Optically Thin
For an optically thick gas it would be very convenient if L, could be
obtained by applying a simple correction factor to the L,,, computed
from equation (5-51). It has been found that a useful technique is to
introduce a correction coefficient C so that

Then the incoming heat flux in equation (5-50) can be obtained as


The coefficient C will now be examined by considering the example

of a radiating gas between infinite parallel plates spaced D apart. Using
equation (5-54) in equation (5-56) gives

d q ~ i =[ I - exp ( - U A C ~ D ) ] ~ A ~ , ~ ~ A

From equation (5-46) the actual flux received is

To see how well these fluxes compare, the ratio

1- ~ E ~ ( c L ~ D )
1 - exp ( - 2 c a ~ D )

is plotted in figure 5-10 using a value of C = 0.9 for a range of optical

thicknesses aAD. This value of C was found by trial to yield a ratio close

to unity for all aAD and hence serves as a useful correction coefficient.
In table 5-1 the mean beam length L,, is given for a number of
geometries along with a value L , that provides reasonably good radiative
fluxes for nonzero optical thicknesses. The values of C are found to be
in a range near 0.9 (refs. 3 to 5). Hence, it is recommended that for a
geometry where L, values have not already been calculated, the approxi-
mation can be used
Le=0.9Le, 0-0.9 - (5-57)

FIGURE 5-10.-Ratio of emission by gas layer to that calculated using a mean beam
length I.,.= 1.80.



Mean beam
Characterizing corrected
Geometry of gas volume dimension for finite

Hemisphere radiating I Radius R

to element at center I
of base
Sphere radiating to its Diameter D
Circular cylinder of Diameter D
height equal to diam-
eter radiating to I
element at center of
Circular cylinder of Diameter D
infinite height radi-
ating to convex
bounding surface
Circular cylinder of Diameter D
semi-infinite height
radiating to element
at center of base
Circular cylitider of Diamrtc~rD
semi-infinite Ilri?~l~t
~aciiatil~g to tantilt.
Circular cylinder of Diameter D
height equal to diam-
eter radiating to en-
tire surface
Cylinder of infinite Radius R
height and semi-
circular cross sec- I
tion radiating to
element at center of
plane rectangular
Infinite slab of gas Slab
radiating to element I thickness D
on one face I
" Cc,rrrctions are those suggested hr Hottel rt sl. (refs. 3 and 4) or Erkert (ref. 5). Correclions w r r r chosm to provide
rnaxinmnl I,, whrrr these referrnces disagree..


Mean beam \lean beam
length for length
Characterizing optical corrected
(;eometry of ~ a volume
s dimension thickness. forfinite C=l.,~/L,,,,,
arL,+ 0 , optical

Infinite slab of gas Slab 20 1.80 0.90

radiating to both thickness D
bounding planes
Cube radiating to a Edge X %x 0.6X 0.90
Rectangular parallele- Shortest edge
radiating X 0.90X 0.82X 0.91
to 1 x 4
radiating .86X .71X .83
l X I X 4 <
radiating .89X .81X .91
to all'faces
to 2 x 6 1.18X ....................................
to 1 x 6 1.24X ....................................
to 1 x 2 1.18X ....................................
to all 1.20X ....................................
, faces V
Gas volume surround-
ing an infinite tube
bundle and radiating
to a single tube:
Equilateral tri- Tube
angular array: diameter
S=2D D, 3.4(S-D) 3.0(S- D ) 0.88
S=3D and 4.45(S-D) 3.8(S- D) .85
Square array: spacing
S=2D between tube 4.1(S- D ) 3.5(S-D) .85
centers S

" Corrections are those suggested by Hottel et al. (refs. 3 and 4) or Eckert (ref. 5). Corrections were chosen to provide
maximum L, where these references disagree.



In furnaces the walls are usually rough and soot covered so they act
practically as black surfaces. An important industrial problem is the
radiant exchange between the furnace gas and the walls. In this section
the simplified case of a black enclosure will be considered. The total
radiation excliange will be considered between the entire gas volun~e
and the enclosure boundary. This development will be carried out by
application of the mean beam length.

5.6.1 Radiation from Gas to All or Portion of Boundary

The mean beam length was found to be approximately independent
of ax as evidenced by equation (5-57). This ineans that L , can be used
as a characteristic dinlension of the gas volulile and regarded as a con-
stant while integrating over wavelength. The total heat flux from the
gas incident on a surface is found by integrating equation (5-44) over A

where L , is independent of A. Now define a gas total emittance E, such

qi = ~ , a T i (5-59)

Equating the last two relations results in

The E , is a convenient quantity which can be presellted in graphical

form for each gas in terms of the variables Le and T,. Then for a particu-
lar geometry and gas condition the E , can be found and applied by use
of equation (5-59).
The 6, charts that will be presented here have been developed by
Hottel (ref. 3) from many experimental measurements. The gas pressure
will enter as a parameter because of the dependence of ax on the gas
density. If the gas is in a mixture, both the pressure of the mixture and
the partial pressure of the radiating constituent under consideration
will be parameters. A chart of eg was given for carbon dioxide ( C O z ) in

figure 1-11. Charts are presented here in more detail for COz and water
vapor (figs. 5-11 and 5-13). Additional charts for sulfur dioxide, ammonia,
carbon monoxide, methane, and a few other gases can be found in
reference 4. The discussion here will be limited to radiation of COz and
water vapor.
When computing an average radiation to area A (a part or all of the
bounding surface) by the relation from equation (5-59)

the mean beam length for the gas geometry is first obtained from table
5-1 or equation (5-57). Then, by knowing the partial pressure of the
gas and its temperature, the gas emittance is found by using figures
5-11 to 5-15. Figure 5-11 gives the total emittance of COz obtained
experimentally using a mixture with air or other gases so that the total
pressure of the mixture was at 1 atm while the partial pressure of the
COr was varied. The dotted lines are regions unsupported by experi-
mental data. For a mixture total pressure other than 1 atm, there is a

Gas temperature, Tg, OR

5-11.-Total emittance of carbon dioxide in a mixture having a total pressure of
1 atm (ref. 3).

.05 .08.1 .2 .3 .5 .8 1
Total pressure of gas mixture, P, atm

FIGURE5-12.-Pressure correction for COr total entittance for values of P other than
1 atrn (ref. 3).

Gas temperature, Tg, OR

FIGURE5-13. -Total emittance of water vapor in limit of zero partial pressure in a mixture
having a total pressure of 1 atm (ref. 3).

pressure broadening correction that is to be applied (ref. 3). This is

given as a multiplying coefficient, Ccoz, in figure 5-12. In the case of
water vapor the emittance is influenced in a slightly more complex man-
ner by both the partial pressure of the water vapor and the total pressure
of the gas mixture. For correlation purposes, the values in figure 5-13 are
e m i t t a n c e ~that were "reduced," by using a factor depending on pH20
and ~ r ) ~ . , ~ toLlimiting
e, values as the partial pressure approaches zero
in a mixture having a total pressure P= 1 atm. A multiplying correction

Average pressure, (P + pH 0)/2, atm


FIGURE 5-14.-Pressure correction for water vapor total enlittance for values of /)~~r,
and P other than 0 and 1 atm, respectively (ref. 3).

(a) Gas temperature (b) Gas temperature (c) Gas temperature

T,= 720" R. T,= 1460° R. T , 3 2160" R.
5-15.-Correction on total emittance for band overlap when both COz and water
vapor are present (ref. 3).

coefficieilt CHeois given in figure 5-14 to account for the actual partial
and total pressures involved. If CO2 and water vapor are both present in
the gas mixture. an additional quantity A6 must be included to accou~lt
for an emittance reduction resulting froni spectral overlap ofthe COe and
H 2 0 absorption bands. This correction is found from figure 5-15. For a
inixture of COz and water vapor in a nonabsorbing carrier gas the emit-
tance is then given by

5.6.2 Exchange Between Entire Gas Volume and Boundary

Hottel (ref. 3) has provided a simple approximate procedure appli-
cable when the cooled enclosure boundary is black and is at a temperature
where it will emit appreciable radiation. The total energy removed at
the wall must equal the energy being supplied by some external means
such as combustion to the gas. Taking a heat balance on the gas then
shows that the net nverc1g.e heat flux being I-enloved at the wall is the gas
emission minus the emission from the wall that is absorbed by the gas,
that is,

The a,(T,,) is the absorptance of the gas for radiation emitted from the
wall at temperature T,. The a,(T,,) depends on T w as this determines
the spectral distribution of the radiation received by the gas. According
to reference 3, the a, can be found from


The E,+,,~ and egZoare respectively cco2and E H ~ Oobtained from figures

5-11 and 5-13 evaluated at the abscissa TI, and at the respective
parameters pco,L,(T,,/T,) and pHzoL,(T,,/T,). For further information
the reader is referred to references 3 and 4.

EXAMPLE: 5-2: A cooled right cylindrical tank 4 ft in diameter and 4 ft

long is filled with hot gas at a total pressure of 1 atm. The interior surface
of the tank is black. The gas is composed of two constituents; a trans-
parent gas at a partial pressure of 0.75 atm, and the remainder carbon
dioxide. The gas is uniformly mixed at a temperature of 2000" R. Com-
pute how much energy must be removed from either end of the tank
to keep it cool if the tank walls are all sufficiently cooled so that only
radiation from the gas is significant.
The geometry is a finite circular cylinder of gas and the radiation to
its base will be computed. Emission from the cooled walls is neglected.
Using table 5-1, the corrected mean beam length for this geometry is
Le=0.60D=2.4 ft. The partial pressure of the C 0 2 is 0.25 atm, so that

From figure 5-11, E ~ ~ , ( ~ ~Tg)=0.13

~ , L , , and Ccoz from figure 5-12
is 1.0 since the mixture total pressure is unity. Assuming the base of
the tank to be sufficiently cool so that its emitted energy is negligible,
the energy to be removed is from equation (5-61)

Qi= E ~ ~ ~ ( =0.13
T T $X~0.173 X 1 0 - X ( 2 0 0 0 ) 4 4 ~ = 4200
5 -11r


The mean beam length approach in the previous section was concerned
with the radiation from a gas volume to all or a portion of a black en-
closure boundary, or the average exchange between the gas and a black
isothermal enclosure. For a more general analysis of radiation in an
enclosure, the exchanges of total radiation must be considered between
the various pairs of bounding surfaces having different temperatures.
This involves integrating the exchange relations involving TA and SAover
all wavelengths. An example requiring such an integration was outlined
in example 5-1 which considered a parallel plate geometry.
A form of the spectral equations was given by equation (5-20) that
relates the gas blackbody emissive power and the spectral fluxes d q ~
supplied in a differential wavelength interval to each surface. In the
solution accounting for spectral effects as described in section 5.3 of
volume I1 for enclosures filled with nonabsorbing media, the set of en-
closure equations is solved at each wavelength for the d q (assume
~ that
the surface temperatures are specified) and the results then integrated
over all wavelengths. For a gas radiation problem the gas properties
vary so irregularly with wavelength that the detailed integration over A

would he a practical impossibility. This leads to a consideration of using

finite wavelength bands.

5.7.1 Band Equations

An approach that can be used to integrate over wavelength is devel-
oped by dividing the spectrurll into absorbing and nonabsorbing bands.
For a typical band of width AX, the integration of equation (5-20) gives

Now it is assumed that the bands are sufficiently narrow that

- -
d q ~j,,EA, j , TA, k-j, a h , /,.-j,e A b , j, and e A b , , can be regarded a s constants
over the band width, being characteristic of some mean wavelength
within the band or in the case of 7 and E being averaged over the band
a s will be described in section 5.7.2. Then equation (5-68) can be written
for band 1 as

I n a spectral region where the gas is essentially nonabsorbing, ? I =1

and & I =0 so that equation (5-69) reduces to

which is of the form of equation (3-19) in volume 11.

5.7.2 Transmission and Absorption Factors

The 5, k - j in equation (5-69) is found from equation (5-27) by taking
an integrated average over the band, that is,

Similarly the is obtained as

(5- 72)
For each small band width

and to evaluate al and ; i l only the single integral is needed


From equation (4-73) the cul can be expressed if desired in terms of the
effective band width as

To obtain 7 and 5 for use in equation (5-69), the integral in equation

(5-75) must be evaluated between pairs of finite surfaces in t h e various
wavelength bands involved. It is evident that, when there are more than
a few bands that absorb appreciably, the solution involves considerable
computational effort. A simplification has been developed b y Dunkle
(ref. 6) that saves considerable labor and yields good accuracy. Dunkle
assumes that the integrated band absorption is a linear function of path
length. This has some physical basis as it holds exactly for a weak band
as shown in equation (4-76a). Also it is the form of some of the effective
band widths in table 4-11. As shown in reference 6 by means of a few
examples, reasonable values of the energy exchange are obtained by
use of this approximation. Hence let a1 in equation (5-73) have the
linear form
ar ( S ) = ClS (5-76)

Now define a mean value of the path length S, called the geonzetric

m e a n b e c ~ mlength Sk-,. This mean length is such that c x l evaluated from

ec~uation(5-76) by using S = S k - j will yield 6 r , k - jas
found from the
integral in equation (5-73). Substituting 6 1 , k - j = C I S ~ - and
j al=ClS
into equation (5-73) gives

Hence the relation to obtain the geo~lletricmean beam length Sk-j is

which is only dependent on geometry. Dunkle (ref. 6) has computed and

tabulated values of 3 k - j for the geometries of parallel equal rectangles,
rectangles at right angles, and between a differential sphere and a rec-
tangle. Results for equal opposed parallel rectangles are sl~ownin figure
5-16. Tabular values for parallel rectangles, and rectangles at right
angles given in tables 5-11 and 5-111. Other Sk-i values are referenced
by Hottel and Sarofim (ref. 4).
For a given gas at uniform conditions, the effective band width correla-

Length of rectangle ,
Distance between rectangles c

Width of rectangle b
Distance between rectangles' c

5-16.- Geometric mean beam lengths for equal parallel rectangles (ref. 6).
[From ref. 61

ti) m
ati) 0
g 2E 9S
ti) *
g gn

ti) m
LC m
V3 -1
% $B
2 2
U3 m
% z2"s
e V3
V3 m
K s9 9r

2 gz
m r-
m m
X s2

* zs
0 *

N m
m n
2 g z
2 3 0 0
30 n
0 PI
6 21
0 4
m o
ti) 3.6
r- m
8 or-

tions discussed in chapter 4 can be used to obtain A / . Using the AX as

described in the next paragraph fixes a / from equation (5-74) and by
fitting with a linear path length dependence determines C I from equation
(5-76). Corresponding to the geometric mean beam length between
surfaces j and k, the 6 can then be found as C I S and the ;i as 1- 6. Then
ecluations (5-69) and (5-70) can be solved for each wavelength band I.
The total energies at each surface X. are found from

The values of AX, the wavelength span of each band, must be specified
in order to carry out the solution. As discussed after equation (4-73),
this span can increase with path length. Edwards and Nelson (refs. 7 and
8) give recommended spans for COz and H z 0 vapor; these values are
reproduced in table 5-IV for the parallel plate geometry. Note that these
values are given in terms of wave number rather than wavelength. For
other geometries, Edwards and Nelson give methods for choosing,
approximate spans for COz and H z 0 bands. Briefly, the method is to
use approximate band spans based on the longest important mass path
length in the geometry being studied. With this in mind, the limits of
table 5-IV are probably adequate for problems involving COz and H z 0



[From refs. 7, 8, and 111

Gas Band, A,
center, 71,
Band limits, 7 ,cm-I

co, 15

H, 0 6.3

"/Iarr found for various hands from tables 4-11 and 4-111. Terms such as /i,r/1.78arr /i/2(1- r , , ) In,m eq. (17)and
tallles 1 and 2 of ref. 7.

If all surface tenlperatures are given in the problem at hand, the results
found from equation (5-78) conlplete the solution. If qk is given for n
surfaces and Tk for the remaining N-n surfaces, then the unknown
surface temperatures are guessed and the calculated qk are compared
to the given values. If they do not agree, then new values of Tk for the
11 surfaces are assunled and the calculation is repeated. This procedure

is continued until there is agreement between given and calculated

q , for all k. For the given T!,, equation (5-21) expressed as a sum over
the wavelength bands gives the required energy input to the gas.
Two example problems will now be presented for an isothermal gas
in an enclosure. Then the discussioll will continue to consider removing
the restriction of uniform conditions in the gas.

EXAMPLE5-3: Two black parallel plates are separated by a dis-

tance of D = 1 m. The plates are of width W = 1 m and of effectively
infinite ledgth (fig. 5-17). The space between the plates i s filled with
carbon dioxide gas at a pressure of 1 atm and a temperature of 1000 K.
If plate 1 is maintained at 2000 K and plate 2 is maintained at 500 K,
find the energy flux that must be supplied to plate 2 to maintain its
As shown by figure 5-17 the geometry is a four boundary enclosure
formed by the two plates and the two open bounding The open

Boundary 3

5-17.-Isothermal carbon dioxide contained between black plates (see example

bounding planes are perfectly absorbing (i.e., nonreflecting) and radiate

no significant energy for the present case as the temperature of the sur-
rounding environment is assumed low. The energy flux added to surface
2 will be found by using the enclosure equation (5-20) where k = 2 and
N = 4. Since all surfaces are black, E A , =
~ 1 and equation (5-20) reduces

The self-view factor F2-2= 0 in this case, and e ~ t , , ~e ~=b , 4 = 0 ,so the
summations can be written out as

To simplify the example, it will be carried out by considering the entire

wavelength region as a single band. Then to obtain the total energy
supplied to plate 2, integrate over all wavelengths to give

Using the definitions of total transmission and absorption factors which

i?-lo~;=I' ?A,z - l e ~ a 1dA

and so forth, the q2 becomes

To determine the ;i and 6 , the concept of geometric mean beam length

will be used. For opposing rectangles, from figure 5-16 a t an abscissa
of 1.0 and on the curve for length to spacing ratio of a,the Sp-llD= 1.34
or 1.34 m. To determine 82-1which determines the emission of

the gas, use the emittance chart in figure 5-11 at a pressure of 1 atm, a
beam length of 1.34 m (4.40 ft), and T,= 1000 K (1800" R). This gives
62-1=0.22. When obtaining 72-1, note from equation (5-79) that the
radiation in the r2-1 term is exb, 1 and is coming from wall 1. Therefore,
it has a spectral distribution different from that of the gas radiation. To
account for this nongray effect, equation (5-65) will be used with E +
evaluated at p c o , S 2 - 1 ( ~ 1 / ~ =
, ) 1.34 (2000/1000) = 2.78 (atm) (m)= 9.11
(atm) (ft) and TI= 2000 K (3600" R). Then using figure 5-11 (extrapolated)
and equation (5-65) results in

From section of volume 11, the configuration factor FZp1

given by

Then Fp-3=Fp-4=$(l-0.414) =0.293.

The (Y2-3= (Y2-4 remain to be found. For adjoint planes as in the geom-
etry for table 5-111, the following expression from equation (12) of
reference 6 can be used, obtained for the present case where b-+ m,
a = 1, and c = 1:

Using figure 5-11 at pS= 0.752 (atm)(m)= 2.46 (atm)(ft) and T,=1800° R
gives ( Y z - ~ = (X2-4 = 0.19. Then

The solution is now complete. Note that the largest contribution to qz

is by energy leaving surface 1 and being absorbed by surface 2. Emission
from the gas to surface 2 and emission from surface 2 are negligible.

EXAMPLE5-4: Parallel nongray plates are 1 in. apart and are at tem-
peratures of TI = 2000" R and Tz= 1000" R. Pure COz gas at 10-atm pres-
sure and T,= 1000" R is between the plates. The plate hemispherical

spectral emissivity as a function of wave number is approximated by

the following table:

Assume that only the 15-, 10.4-, 9.4-, 4.3-, and 2.7-pm Con bands
cause significant attenuation in the gas. Compute the total heat flux being
added to plate 2.
In example 5-1 the spectral exchange was found for radiation between
infinite paralleI plates with a gas between them. The total energy added to
plate 2 is found by integrating equation (5-24b) over all wave numbers

In this example =~ 0 and TI,= T2 so the

, ~ 9.1 simplifies to

The integration can be expressed in finite difference form as a sum over

wave number bands. For the Ith band let = € 1 , T s = 71, and so forth.
~ bl ( )T z ) ] I A ~ I
~ ; ~ l [ e b (- e
9'=-C I

From equation (5-74) the 71 can be written as

where the is the integrated band width which includes the integrated
path length variation for a parallel plate geometry. The 9.1 now becomes

The needed quantities and results for this problem are shown in the
table. Values of 4 were computed from the exponential wide band
correlation data of tables 4-11 and 4-111 using the mean beam length
from table 5-1 as the effective path length. The wave number spans
Aq1 were computed from the data of table 5-IV, and values of

for the nonabsorbing regions were con~putedusing the F O - X T factors

from table V of the appendix to volume I. When the band correlations
give A1 > Aqr, then A I / A ~1.0
~ =is used since physically A! cannot
exceed AT!.

[el,(TI)- e b ( T z ) l ~ A ~ l ,
Band, q , cm-' Btu

0 to 556
556 to 778 (15 pm)
778 to 849
849 to 1013 (10.4 p m )
1013 to 1141 (9.4 p m )
1141 to 2221
2221 to 2430 (4.3 p m )
2430 to 3573
3573 to 3750 (2.7 pm)
3750 to m

The result for q z compares with a value of -9371 Btu/(hr)(ft2) found for
the same problem by Edwards and Nelson (ref. 7).8 They use the network
method of Oppenheim (ref. -9) in deriving the energy transfer equation,
which, of course, gives the same result as that used here. Partial emit-
tances were used in place of the band correlations for computing gas
properties, and these led to slightly different wave number spans for
the bands used in reference 7.


Edwards and coworkers (refs. 10 to 12) have further extended the
band and geometric mean beam length approaches to account for non-
isothermal gases. Removing the isothermal gas restriction introduces
considerable additional complication. In a nonisothermal case, the band

The reslllts of rrf. 7 l ~ a v ran rrror in q,,, for the 2430- I<,' range. Thr cornparisc,n dr,srribed here is aftpr
correction nf that rrror.

absorption may vary strongly with position in the gas. Then, a linear ab-
sorption law may be valid in one portion of a gas, but a power law might
be necessary in another portion. The Curtis-Godson technique that will
be discussed in section 5.8.1 is the basis for one engineering treatment of
nonisothermal gases (refs. 10 and 13 to 16). Another treatment, due chiefly
to Hottel and coworkers (refs. 3, 4, and 17), is the zoning method dis-
cussed in section 5.8.2. The methods of chapter 3 can be used in simple
geometries to treat radiation in nonisothermal gases. In connection with
this, the exchange factor approximation is given in section 5.8.3. The
methods in this chapter and the Monte Carlo techniques of chapter 6 are
powerful enough to treat multidimensional problems.

5.8.1 The Curtis-Godson Approximation

An accurate and useful method for solving thermal radiation problems
in nonuniform gases is the Curtis-Godson approximation (refs. 10 and 13
to 16). In this method, the transmittance of a given path through a non-
isothermal gas is related to the transmittance through an equivalent
isothermal gas. Then the solution can be obtained by using isothermal
gas methods. The relation between the nonisothermal and the isothermal
gas is carried out by assigning an equivalent amount of isothermal absorb-
ing material to act in place of the nonisothermal gas. The amount is
based on a scaling temperature and a mean density or pressure that is
obtained in the analysis. These mean quantities are found by specifying
that the transmittance of the uniform gas be equal to the transmittance
of the nonuniform gas in the weak and strong absorption limits.
Goody (ref. 15), Krakow et al. (ref. 13), and Sinlmons (ref. 16) have
discussed the Curtis-Godson ~llethodfor the case of attenuation in a
narrow vibration-rotation band. Excellent comparisons with exact
numerical results were obtained. Weiner and Edwards (ref. 12) have
applied the method for engineering environments, that is, for steep
temperature gradients in gases with overlapping band structures. Com-
parison of the analysis with experimental data was again excellent. In
the following development of the use of the method, spectral variations
will be expressed in terms of wave number -q= 1 / X since the absorption
band correlations are often expressed in terms of this variable. The
Curtis-Godson technique is most useful when the temperature distribu-
tion in the gas is specified. If the gas temperature distribution is not
known, an iterative procedure would have to be developed for its deter-
mination. This is not considered here as the method is not too practical
for that type of calculation.
For a nonuniform gas the absorption coefficient U Q i s variable along
the path. An effective band width &(s)is defined i n this instance
analogous to equation (4-73) but using an integrated absorption coefficient

Al(S) = Inbsoliltioll [-I:

h n l ~ dwidth (1 - exp a n ( q , s * ) ~ s * ] dr)

=A?,-L [-I," {exp an(?, s * ) ~ s * ] ]h (5-81 )

Similarly for a path length extending from S* to S , the effective band

width is

The equation of transfer will now be placed in a form utilizing&(s) and

The integrated form of the equation of transfer for intensity at S as a
result of radiation traveling along a path from 0 to S is given from equation
(3-1) by

i ; ( q i S ) = i ; ( q , 0 ) exp [loS an(7,s * ) ~ s * ]

+%an(?, ~ * ) i ; h (s*)
~ , exp [-[*an(q,
S * * dS** dS* (5-83)

Now note that

{I-.., [-I* a,(,, s**)ds**]]

Insert equation (5-84) into equation (5-83) to obtain

Equation (5-85) is now integrated over the band width Aql of the lthband,
and the order of integration is changed on the last term. It is assumed that
i ; ( q , S ) , i;(r], 0 ) , and ikh(q,S ) can be approximated by average values
within the band. Then

~ ; ( s ) A v I =i ; ( ~ )I l {exp [ - y n0 q c g , S * ) ( / S * ] } ~ ~

, { I - exp [-[;*</,q(r), ~.")ds*"] J ( l ~ ~ l s * (5-86)

Ec~uations(5-81) and (5-82) are substituted into equation (5-86) to obtain

the equation of transfer in terins of the 4

An alternate forin can be found by integrating equation (5-87) by parts

to give

Equations (5-87) and (5-88) are nearly exact forins of the integrated
equation of transfer in terms of the band properties. The only approxima-
tion is that the intensity in each term does not vary significantly across
the wave number span of the band.

Note that, for a uniform gas, equation (5-88) gives, since di;, I,/dS=O,

where the u subscript denotes a uniform gas.

In order to compute ii (S) or i;,,,(S) from equation (5-87), (5-88), or
(5-89), expressions are needed for the effective band width 21for non-
uniform and uniform gases. From equations (4-76a) and (4-76b), the
limiting cases of 4 for weak or strong absorption in a uniform gas have
the form
At. ,,(S) =el, ,PIS',, (weak) (5-90a)

21,.(S) = Cz, lpUSY2 (strong) (5-90b)

where the C,,l and C2,1 are coefIicieilts of proportionality for the ltll band.
For the nonuniform gas the effective band width will depend on the
variation of properties along the path. The effective band widths are

then obtained by applying equations (5-90a) and (5-90b) locally along

the path. This gives for a weak band

A ~ ( s ) = c ~ , ~ I ~ ~ ~ ( s * )(weak)
~s* (5-91 a)

where the p is a function of position S * along the path. Similarly for the
strong band by first squaring equation (5-90b)

q ( s ) = c:,, ls p2(~*)d~*
so that
K,(s) = C2,r [I0' p2(s*)ds*] ' I 2 (strong) (5-91 b)

It has been assumed that the Cl,l and C2,, do not vary along the path.
In the Curtis-Godson method the nonuniform gas is replaced by a n
effective amount of uniform gas such that the correct intensity is obtained
rrt the weak a n d strong absorption limits. To have the uniform intensity
equal the nonuniform intensity, equate the results from equations (5-89)
and (5-88) to obtain

i;(O) [ A ~ I - L ~ I , ~ , (-ti;,*,
S ) I uAl,~(5')
cli;, ,(S*)
=i; (0) [A~~--X,(S)I i;,,(o)Jl(s) + I S ~ I ( ~ - ~ *ds* ) d s*

which simplifies to

To have equation (5-92) valid at the weak absorption limit, substitute

4,,, from equation (5-904 and from equation (5-91a) to obtain the
following after canceling the Cl, 1:

Similarly at the strong absorption limit, insert equations (5-90b) and

(5-91b) into equation (5-92) to obtain

For a known distribution of temperature and density in a nonuniform

gas, equations (5-93a) and (5-93b) can be solved simultaneously for
p,, and St, which are the equivalent uniform gas density and path length
for that particular band. The i;, h , ,,(Ti,) is not an additional unknown
since the temperature Ti, corresponds to p,,. Then equation (5-89) can
be used for any effective band width dependency on p,, and S,, (i.e., not
at only the weak and strong limits) to solve for i;, ,,(S). This will exactly
equal the intensity i; (S) in the nonuniform gas in the weak and strong
li~llitsand will usually be a good approximation for intermediate absorp-
tion values. Once the intensities are found, the heat transfer can be
obtained by using the relations for a uniform gas. All~lostillvariably the
evaluation of equations (5-93a) and (5-93b) will require numerical
integration. Because the Curtis-Godson method requires evaluation of
at least two integrals for each band along each path, it may i n nlany cases
be equally feasible to evaluate the exact equation (5-87) o r (5-88). This
is especially true if the problem is to be solved by electronic computer.
As originally fornlulated (see, for example, the discussion in Goody
(ref. 15)),the Curtis-Godson approximation was limited to application over
a small frequency span in an absorption band. The limitation was due to
considerations of line overlapping, and the change i n the spectral
position of important lines with temperature. It has been shown, however
(see, for example, Wiener and Edwards (ref. 12) and P l a s s (ref. 18)),
that the method gives good results even when applied to situations with
large temperature gradients while using fairly wide frequency spans.
These references also account for overlapping absorption bands.
The Curtis-Godson technique appears to have application even in
nlultidimensional problems, even though it was originally applied to
one-dimensional atmospheric problems. Although no one has explictly
carried out such calculations, it should be possible to proceed a s follows.
For a known field of temperature and density, the medium and the
boundaries are subdivided into convenient nearly isothermal zones.
Between each two zones, an equivalent uniform path length and density
are found for each important band by the use of equations (5-93a) and
(5-93b). Based on these parameters, the values of A, can b e obtained from
one of the correlations of gas properties. The uniform gas analysis of sec-
tion 5.7 can then be carried through to obtain intensities a n d heat flows.

5.8.2 The Zoning Method

The zoning method consists of subdividing nonisothermal enclosures
filled with nonisothermal gas into areas and volumes that can be con-
sidered essentially isothermal. An energy balance is then written for
each division of area and volume. This leads to a set of simultaneous
equations for the unknown heat fluxes or temperatures in the same
manner as the procedure discussed in section 5.3 for an isothermal gas.
The method is not elegant in a formal mathematical sense but is practical
and very powerful. Hottel and Sarofim (ref. 4) discuss the method at
some length. Applications in multidimensional situations have been
carried out by Hottel and Cohen (ref. 17) and Einstein (refs. 19 and 20).
The discussion in this section is limited to cases when the energy
exchange is only by radiation; extension to situations including conduc-
tion and convection is found in chapter 7 and reference 4.
The zoning method has an advantage over the Curtis-Godson method
outlined in section 5.8.1 because unknown temperature distributions in
the gas can be treated. The Curtis-Godson technique is most useful
where the temperature distribution is known; if the distribution is not
known, some method of iteration on the gas temperature must be
The basic concepts of the zoning method will now be developed for a
gas with a constant absorption coefficient. Consider a volume Vy as in
figure 5-18 and a surface Ak. From equation (1-33) the emissive power
from a volunle element ~ V isY LE.rraxi;,dVvdA or per unit solid angle
around dVy it is a ~ i i ~ d V Y d AThe
. surface element dAk subtends the
solid angle clAk cos /3k/S$-k when viewed from dVy. The fraction of
radiation transmitted through the path length SY-k is

Multiplying these factors together and integrating over VY and A,< gives
the spectral energy arriving at surface Ak from a gas volume VY a s

x exp [Is: I
U A ( ~ * ) ddAkdVy
~* (5-94)

If aA(y) is assumed uniform, then the exponential factor becomes


Path direc-
tion S

Location Sc

F I G ~ R E5-18.-Radiation from gas volume VY to area Ak.

exp [- ax(S,<- S Y ) I =~ ~ ( S Y - P )
The entire gas volume has been divided into finite subvolumes VY and
the assumption is made that conditions are uniform over each VY.Equa-
tion (5-94) then simplifies to

cos p,
&hi, Y - ~ P = dAaxi;,(y) T ~ ( S Y d- A~ ~) d V y (5-95)

If the gas in addition is gray, then ecruation (5-95) can be integrated

over all wavelengths to obtain the total incident energy on Ax. as

p, cos
.i(~y-*)d~,d~v (5-96)

Now define the gas-surface direct exchange area as


Equation (5-96) then can be written as

Thus the energy arriving at Ak, q;,y-jiAk is the blackbody emissive

power of the gas in V Y ,aT$ radiated from an effective aream.
Let the entire gas volume be divided into finite regions. The energy
flux incident upon surface element Ak from all r of the gas volume
regions is then

Now consider the interchange between the bounding areas of the

enclosure. The energy leaving surface area Aj and reaching Ak is, for a
nonisothermal gas with uniform gray properties,

where, as in the usual enclosure theory, q,,j is uniform over Aj. Now
define the surface-surface direct exchange area as

Sjsk IAj cos
P j cos P k

Equation (5-100) can then be written as

qi, j-kAk=~jSk 90,j

Thus the energy from Aj arriving at Ak, qi,j-kAk, is the energy flux
leaving Aj times an effective area S F . The energy flux incident upon
area Ak as a result of the fluxes leaving all N surfaces of the enclosure
is then

Now the total energy flux incident upon surface Ak can be obtained as

T h e usual net radiation equations (eqs. (3-1) and (3-2) of vol. 11) also

apply at surface A!,.

qk=qo, k-qi,k (5-105)

For problems where T y is given for all gas volunle elements P7Y, equa-
tions (5-104) to (5-106) are sufficient to solve for N unknown values of
either Tk or qk, or some combination of Nvalues of Tk and qk. The other N
values of Tk and qk must be provided as known boundary conditions.
The methods of section 3.3 of volume I1 can be directly applied. Values of
sjsk and g m . have been tabulated for cubical isothermal volumes and
square isothermal boundary elements by Hottel and Cohen (ref. 17).
Hottel and Sarofim (ref. 4) present a reference table to factors for eleven
other geometries, and also an extensive tabulation of the factors for the
cylindrical geometry.
When the T y of the r gas elements are unknowns, then r additional
equations must be found. These are obtained by taking an energy
balance on each gas zone. In radiative equilibrium, for each gas element
VY the emission and absorption of energy are equal. Then for a gray gas
with uniform properties a heat balance on the volume region VY gives

It is assumed that a is uniform throughout the enclosure, and that

V y and all the Vy* are each isothermal. As usual in the enclosure calcu-
lation methods, q,, k is taken as constant across A k .
Define the surface-gas direct exchange urea as

Comparing equation (5-108) with equation (5-97) shows that there is


reciprocity between the surface-gas and gas-surface direct exchange

- -
S I ~ =~ ~Y Y S I ~ (5-109)

Now define the gas-gas direct exch.ange area as

Substituting equations (5-108) to (5-110) into equation (5-107) gives

The gy*gy have also been tabulated (ref. 17) so that equation (5-111)
written for each VY provides the additional set of T equations required
to o b t a i ~the gas temperature distribution.
The notation developed by Hottel and coworkers has been used in
this section with only slight modification in developing the preceding
equations. A comparison with the derivations of section 5.3 shows that,
in terms of the notation used there (eq. (5-9)), the following identity
Fj-k?j-kAj = J k (5- 112)

The gas-absorptance factor in equation (5-lo), Fj-kcj-kAj is gener-

ally not related in a useful way to gvsk. The latter quantity is derived for
an element of gas volume, while Fj-k&j-kAj is concerned with the entire
gas volume.
Hottel and coworkers (refs. 3 , 4 , and 17) have developed the approach
outlined in this section even further. Allowance for spectral variations
in gas properties, done in an approximate but easily carried out manner,
is possible. Variations in properties with position in the enclosure are
handled by defining a suitable mean absorption coefficient between each
set of zones. Einstein (refs. 19 and 20) modified the @ and @- factors to
give better accuracy when strong gradients are present. All of these ap-
proximations become difficult to carry through if the absorption
coefficient is a strong function of temperature.

5.8.3 The Exchange Factor Approximation

The specific situations discussed here are regions of nonisothermal
gray gas between parallel plates, concentric cylinders, or concentric

spheres. These geometries have been considered in some detail in

chapter 3. The purpose here is to reveal how the engineering concepts
of exchange factors can be applied to extend the results for one of the
previous solutions to more general cases. Specifically it will be shown
how results for black bounding surfaces call be extended to the case
where the surfaces are diffuse-gray. If the walls are diffuse-gray with
henlispherical total emissivity E ( T ) ,no great difficulty is involved in
formulating the governing radiative integral equations as i11 chapter 3.
The boundary condition for the solution of the equation of transfer is that
the intensity leaving a diffuse wall is q , / r where q, is the outgoing
radiation flux previously discussed in regard to enclosure theory. How-
ever, a better approach than solving the integral equations is that of
Perlmutter and Howell (ref. 21) who have shown that, once the results
are available for an analysis with black boundaries, the diffuse-say
wall results can be obtained from simple algebraic relations.
The theory follows the same general development a s the net radia-
tion method (section 3.3.1 of vol. 11). A heat balance at surface A k gives

The energy flux leaving A k is composed of emitted and reflected energy

If q i , I; is eliminated froin ecluations (5-113) and (5-114), the result is

The q i , I; in equation (5-114) call be found in terms of Hottel's exchange

areas as in equation (5-104). However, here we choose to define a dif-
ferent quantity called the exch,ange factor F j - k . The exchange factor
Fj-I;is defined as th,e fraction of the energy leauirzg surface j that is in,ci-
dent 072 surface k when all boundaries are black an,d the interuening
rrredium is i n radiative equilibrium (that is, the heat transfer in the gas
is only by radiation
without any heat sources or sinks). When the gas is
transparent, Fj-,,. becomes identical to the configuration factor Fj-k
(section 2.4.3 of vol. 11). Because the gas is in radiative equilibrium,
energy conservation requires that energy leaving surface 1 niust finally
reach other enclosure surfaces or return to surface 1. Any energy ab-
sorbed in the gas must be reemitted by the gas to n~aintainequilibriuin
and the Fj-,r,includes
. all interactions with the gas by means of which
energy leaving Aj arrives at AI;.

For a general enclosure of N surfaces surrounding a gas in radiative

equilibrium, the incident energy on surface k can be written in terms of
exchange factors as
Qi, k= I=
j= 1
~ oj F j, - k (5- 116)

Note that using exchange areas as in section 5.8.2 would require an

additional term to account for energy emitted by the gas and reaching
the wall. This term is included by definition within the F j - k .
If consideration is restricted to an enclosure having only two surfaces,
equation (5-114) can be written by using equation (5-116) to eliminate
q i , k as

Because the medium is in radiative equilibrium, all energy leaving a

given surface must finally reach an enclosure surface. It follows that


Note that there is no energy being supplied to the gas by any external
means such as combustion.
Because F j - / i is defined as the fraction of energy leaving A j that ar-
rives at AI, for black boundaries enclosing a gas, it can be obtained from
the black-walled solution, which it is assumed has already been found.

pjk ( )
A j v T ? black =(AkaT: -Qk =j - , (5-119)
slof(~ces sltrf(1ces

where the notation + j - k , b is used to emphasize that this is a quantity ob-

tained from the black solution. For a transparent gas F j - k becomes equiv-
alent to the usual geometric configuration factor F j - k for interchange
between two diffuse surfaces. The F factors are found from equation
(5-119) and by using the relations in equation (5-118). Then equations
(5-117) are solved simultaneously for the Qo's and these are used in
equation (5-115) to find Q k . Thus the solution for gray walls can be found
quite simply from the solution with black walls. The procedure will now
be outlined for the infinite parallel plate case.
In the infinite parallel plate case and because the gas absorption

coefficient has been assumed constant. the fraction of energy leaving

surface 1 that reaches surface 2 or -FI-UIIUS~ be equal to the fraction
going from surface 2 to surface 1 or F2-].This arises from the symmetry
in radiation paths experienced by energy leaving either surface. In
radiative equilibrium with no heat sources in the gas, the radiation ah-
sorbed at a position must be reemitted at that position. The radiation
leaving either plate will undergo the same absorption-emission history
while traveling to the other plate.
If F ~ is- found
~ from the black solution as

then Fr-l=Fl-z and for simplicity call them $,). From equation (5-118)
FI- 1
= F z - r = 1- $ b . Equations (5-117) then become

Solving simultaneously for q,, 1 and q,, 2 yields the symmetric relations,

T + I ~ ~ ) $ u ~ T ~ + ~(5-120~)
~ ( ~ -
qo, 1 =
I,bh(€1 + €2 +
- 2 ~ 1 ~ 2 )€ 1 ~ 2

The q,,] is substituted into equation (5-115) to ~ i e l dafter rearrangement

Equation (5-121a) can be written in the alternate form


Evaluating equation (5-121b) for the black case, El = E2 = 0 shows that

Equation (5-121b) gives the energy supplied to surface 1 and removed

from surface 2. If the gas absorption coefficient is independent of tem-
perature, then energy leaving a boundary and reaching a given point in
the gas will be attenuated by the same amount regardless of the gas tem-
perature distribution. Further, any portion of the energy absorbed along
the path is balanced by isotropic emission at each point. Using these
facts, a synthesis of black-wall .enclosure solutions and surface-gas
element exchange factors can be used to find the temperature distribu-
tion in the gas for gray walls.
The energy emitted by a local volume element of gas of area A and
thickness dx between the parallel plates (fig. 5-19) is given by
Q , = 4acrT4(x)Adx

5-19.-Energy quantities for gas between infinite parallel gray plates.

For radiative equilibrium, this must equal the heat absorbed by the vol-
ume element which can be written as

T4(x) - Ti -
1 (qO,lql-d, q0,2dFz-dr
'= Tj-Ti 4aa(Tj-Tj) + dx

Here dFj-dz is the fraction of energy leaving boundary surface Aj that

is absorbed in volume element Adx when the boundaries are black. Again,
the F factors include the energy absorption and reemission from the gas
as the energy travels from the surface to the volume element. Because
radiative equilibrium is the condition being studied here, no energy is
lost during these processes since all absorbed energy at a location must
be re-emitted.
When the entire system is isothermal, equation (5-122) reduces to

This relation is used to eliminate dF2-ds from equation (5-123) written

for the black surfaces. The resulting equation is solved for d ~-dx1 g m n g

where from equation (5-123)

Then by substituting equation (5-125) into equation (5-124)

Substituting equations (5-125), (5-126), (5-120a), and (5-120b) to elimi-

nate & l - d z , dFz-dZ, go, and from equation (5-123) results, after
much manipulation, in the following gas temperature distribution:

Equations (5-121) and (5-127) relate the energy transfer and fourth-
power temperature distribution for the case of a gray gas between gray
walls to the case of a gray gas between black walls for the geometry of
infinite parallel plates. Similar relations for the geometry of infinitely
long concentric cylinders are given in references 21 and 22, and these
plus the relations for concentric spheres (refs. 23 to 25) are given in
table 5-V.

EXAMPLE 5-5: A gray gas of absorption coefficient 0.5 cm-' is con-

tained between gray parallel plates spaced 2 em apart. Plate 1 has
temperature TI= 1000 K, while plate 2 is at TZ = 840 K. The plates have
emissivities of €1 = 0.1 and e2 = 0.2, respectively. What is the energy
transfer between the plates and the temperature of the gas at a point
0.5 cm from surface I ?
If the walls were black, figure 2-6(b) gives, for aD = 0.5 x 2 = 1.0,

so that from equation (S121c)

$ 1 , ~ 0.56
From table 5-V,

and, for this example,

so that

Geometry Relationsa
Infinite parallel plates
I1+ $,)El
l * ' 1E2 + + 2
-Yx -
. .. 1
. .


PIX) = 'f'b(x) + E2db

(E2+ El)$+ 1
(Cb and 9 b are given i n ref. 3 of chapter 2
and figs. 2-6(a) and (bl)

Infinitely long concentric cylinders

( 1 + El)db

Pb(r) + E~ 5%
6 E2 + E$db + 1

(db and p b are given i n refs. 21 and 22)

Concentric spheres


[(;r 11 + El)db

~2 + ~ l ] d b 1 +

y r ) + E2@J rb

9(r) =

~2 ~ l ] @ b 1
+ +

ldb and pb are given i n refs. 23 to 25)

a~efinitions:Ev(l - E,,,~)/E,,,~, 1= Q,/'~~A~IJ($~ - G2)], (ab = Q ~ ~ / [ A ~ O ( $ -, ~t2],

910 = [+lo - ~ d / ( $-,$,J.

The temperature at the point x= 0.5 cm can be calculated from the

result in table V-5

From figure 2-6(a), at the abscissa K / K D = an/aD = 0.25 and on the curve
for ~ g = a D = 1

so that


T4(0.5)=Tj 0.345(Tf - T:) = [0.5+0.345(1- 0.5)lX 10"=0.673 X 1012

which gives

Note that for gray walls a curve of p against a x / a D will only have an
antisymmetrical shape about a.x/aD = 0.5 as in figure 2-6(a) when € 1 = €2.

1. SIEGEL,ROBERT;AND HOWELL,JOHNR.: Thermal Radiation Heat Transfer. I-The
Blackbody, Electromagnetic Theory, and Material Properties. NASA SP-164, Val.
I, 1968.
2. HOWELL, JOHNR.; AND SIECEL,ROBERT:Thermal Radiation Heat Transfer. II-Radia-
tion Exchange Between Surfaces and in Enclosures. NASA Sf-164, Vol. 11, 1969.
3. HOTTEL,H. C.: Radiant-Heat Transmission. Heat Transmission. Third ed., William
H. McAdams, McGraw-Hill Book Co., Inc., 1954, ch. 4.
4. HOTTEL,HOYT C.; AND SAROFIM,ADEI, F.: Radiative Transfer. McGraw-Hill Book
Co., Inc., 1967.
5. ECKERT,E. R. G.; AND DRAKE,ROBERTM., JR.: Heat and Mass Transfer. Second ed.,
McGraw-Hill Book Co., Inc., 1959.
6. DUNKLE,R. V.: Geometric Mean Beam Lengths for Radiant Heat-Transfer Calcula-
tions. J. Heat Transfer, vol. 86, no. 1, Feb. 1964, pp. 75-80.
7. EDWARDS, D. K.; AND NELSON,K. E.: Rapid Calculation of Radiant Energy Transfer
between Nongray Walls and Isothermal H10 and C 0 1 Gas. J. Heat Transfer, vol. 84,
no. 4, Nov. 1962, pp. 273-278.

8. EDWARDS,D. K.: Radiation Interchange in a Nongray Enclosure Containing an Iso-

thermal Carbon-Dioxide-Nitrogen Gas Mixture. J. Heat Transfer, vol. a,no. 1,
Fcb. 1962, pp. 1-11.
9. OPPENHEIM,A. K.: Radiation Analysis by the Network Method. Trans. ASME, vol. 78,
no. 4, May 1956, pp. 725-735.
10. EDWARDS,D. K.; AND WEINER,M. M.: Comment on Radiative Transfer in Non-
isothermal Cases. Combustion and Flame, vol. 10, no. 2, June 1966, pp. 202-203.
Heat Transfer in Non-isothermal Nongray Gases. J. Heat Transfer, vol. 89, no. 3,
Aug. 1967, pp. 219-229.
12. WEINER,M. M.; AND EDWARDS,D. K.:Non-isothermal Gas Radiation in Superposed
Vibration-Rotation Bands. J. Quant. Spectrosc. and Radiat. Transfer, vol. 8, no. 5,
May 1968, pp. 1171-1183.
L.: Use of the Curtis-Godson Approximation in Calculations of Radiant Heating by
Inhomogeneous Hot Gases. Appl. Opt., vol. 5, no. 11, Nov. 1966, pp. 1791-1800.
14. SIMMONS,F. S.: Band Models for Non-isothermal Radiating Gases. Appl. Opt., vol. 5,
no. 11, Nov. 1966, pp. 1801-1811.
15. GOODY,R. M.: Atmospheric Radiation. Vol. I-Theoretical Basis. Clarendon Press,
Oxford, 1964.
16. SIMMONS,F. S.: Application of Band Models to Inhomogeneous Gases. Molecular
Radiation and its Application to Diagnostic Techniques. R. Goulard, ed., NASA
TM X-53711, 1968, pp. 113-133.
17. HOTTEL,H. C.; AND COHEN,E. S.: Radiant Heat Exchange in a Gas-filled Enclosure:
Allowance for Nonuniformity of Gas Temperature. AIChE J., vol. 4, no. 1, Mar.
1958, pp. 3-14.
18. PLASS, GILBERTN.: Radiation from Nonisothermal Gases. Appl. Opt., vol. 6, no. 11,
Nov. 1967, pp. 1995-1999.
19. EINSTEIN,THOMASH.: Radiant Heat Transfer to Absorbing Gases Enclosed between
Parallel Flat Plates with Flow and Conduction. NASA TR R-154,1963.
20. EINSTEIN, THOMASH.: Radiant Heat Transfer to Absorbing Gases Enclosed in a
Circular Pipe with Conduction, Gas Flow, and Internal Heat Generation. NASA
TR R-156, 1963.
21. PERLMUTTER, M.; A N D HOWELL,J. R.: Radiant Transfer Through a Gray Gas Between
Concentric Cylinders Using Monte Carlo. J. Heat Transfer, vol. 86, no. 2, May 1964,
pp. 169-179.
22. GREIF, RALPH; AND CLAPPER,GEANP.: Radiant Heat Transfer Between Concentric
Cylinders. Appl. Sci. Res., Sec. A, vol. 15, 1966, pp. 469-474.
23. RHYMING,R. L.: Radiative Transfer Between Two Concentric Spheres Separated
by an Absorbing and Emitting Gas. Int. J. Heat Mass Transfer, vol. 9, no. 4, Apr.
1966, pp. 315-324.
24. SPARROW, E. M.; USISKIN,C. M.; AND HUBBARD, H. A.: Radiation Heat Transfer in
a Spherical Enclosure Containing a Participating, Heat-Generating Gas. J. Heat
Transfer, vol. 83, no. 2, May 1961, pp. 199-206.
25. VISKANTA,R.; AND MERRIAM,R. L.: Heat Transfer by Combined Conduction and
Radiation between Concentric Spheres Separated by Radiating Medium. J. Heat
Transfer, vol. 90, no. 2, May 1968, pp. 248-256.
Chapter 6. The Monte Carlo Technique
The Monte Carlo technique is a method of statistical sampling of
events to determine the average behavior of a system. In chapter 6 of
volume 11, the method was applied to radiative transfer between surfaces
without an intervening medium. The information given in that chapter
is a necessary prerequisite to that presented here. Based on the model
of radiative transfer in volume 11, extensions are made in this chapter
to the cases of absorbing-emitting media. The model consists of follow-
ing a finite number of energy bundles through their transport histories.
The radiative behavior of the system is then determined from the average
behavior of a number of these bundles.
Monte Carlo has more obvious utility in solving problems of radiative
transfer through absorbing-emitting media than for surface radiation
interchange problems. This is because a complete definition of the local
radiation balance in a gas or other absorbing-emitting media requires
an integration of the incoming radiation, not only from the surrounding
surfaces, but from all volume elements of the surrounding medium.
Such problems are difficult to solve analytically. As described in other
chapters of this volume, much effort has been expended in attempting
to develop standard analytical solution methods. This is often done by
making as many assumptions, reasonable if possible, as are necessary
to obtain an answer and philosophically accepting the resulting loss of
accuracy, if not validity. Surfaces that are black, gray, diffuse, or spec-
ular, and gases that are optically dense, almost transparent, gray, or
isothermal are typical assumptions that fall into this category. Few prob-
lems in radiative transfer are solved analytically without explicitly or
implicitly making one or more of these assumptions, which may or may
not apply to the problem under consideration.
By extending the Monte Carlo model of radiative energy exchange
outlined for surface interaction problems in chapter 6 of volume 11, it
is possible to account for a large variety of effects in gas radiation prob-
lems. This can be done without resorting to the simplifying assumptions
that are often necessary in the analytical approaches as typified by
references 1 to 4.

a absorption coefficient

separation distance between parallel plates

emissive power
fraction of total blackbody emission in spectral region 0-A
volume increment index
number of volume increments
dimensionless path length, LID
path length to absorption
total number of Monte Carlo bundles per unit time
bundle index
probability density function
increment index
energy per unit time
internal energy source rate per unit volume
energy flux; energy per unit area per unit time
randomly chosen number in range 0 to 1
radial coordinate
coordinate along path of radiation (will not have a subscript)
number of events at some position per unit time (will always
have a subscript to avoid confusion with path length
absolute temperature
energy carried by sample Monte Carlo bundle
dimensionless distance, x/D
distance normal to surface
cone angle (measured from normal of area)
dimensionless temperature, TjITI
circumferential angle
optical thickness, aD
Stefan-Boltzmann constant

b blackbody
e emitted
i inner
j volume increment j
1 path length
m ax maximum
min minimum
0 original, or outer
P Planck mean value

dV elemental volume dV
LU wall
1,2 surface 1 or 2
P for cone angle
6 for circumferential angle
h spectrally dependent
* dummy variable of integration


The additional factor introduced to the model previously discussed
in volume I1 is the path length traveled by an individual energy bundle
before it is absorbed or leaves the system. The required relations are
given in table 6-1 as related to a random number (see also example 6-1).
It is possible to allow for variations in gas properties along the bundle
path; indeed, it is in principle possible to account for variations in the
refractive index of the medium by causing the bundles to travel curved



Phenomenon Variable Relation

I Emission from a volume Cone angle p cos p = l - 2 R B

I element with absorption
coefficient ah Circumferential 6= 2.rrRs
angle 0

Wavelength A
Gray gas:

Nongray gas:

Absorption by gas with Path length 1

absorption coefficient ah Uniform gas
I gas properties:

If a problem is solved in which radiative equilibrium can be assumed,

then whenever a bundle is absorbed in the medium, a new bundle must
be emitted from the same point in the medium to ensure no accumulation
of energy. The functions required for determination of the angles and
wavelengths of emission are shown in table 6-1. The new emitted bundle
in the medium may be considered as merely the continuation of the
history of the absorbed bundle, and the history is continued until the
energy reaches a bounding surface.
Under conditions of radiative equilibrium, the total energy Qe emitted
by a volume element dV, is given by equation (1-34), integrated over all
A and not including induced emission

For equilibrium the energy contained in the bundles emitted by the

volume must be equal to the energy contained in the bundles absorbed,
dQe= ~ S d * i (6-2)

where w is the energy per bundle and the number of bundles
absorbed per unit time in dV. Then, if we note from equation (2-19)

where U P is the Planck mean absorption coefficient, equation (6-3) can

be substituted into equation (6-1) to eliminate the integral. T h e n equating
equations (6-1) and (6-2) gives

This allows determination of the local temperature in the g a s from the

gas properties and the Monte Carlo quantities found in the solution. If
u p depends on local temperature T d f ~an, iteration is required. A tem-
perature distribution is assumed for a first iteration to obtain the bundle
histories. These Monte Carlo quantities are used in equation (6-4) to
obtain a new temperature distribution which is then used for t h e second
iteration. The process is repeated until the temperatures converge.
There are so many variations possible on the Monte C a r l o model,

many of which might lead to increased efficiency, that they cannot all
be mentioned here. One of the most frequently suggested is the frac-
tional absorption of energy when a bundle reaches a surface of known
absorptivity. Using such a scheme, the bundle energy is reduced after
each reflection. The bundle history is then followed until a sufficient
number of reflections have occurred to reduce the bundle energy below
some predetermined level. This level is chosen so that the effect of the
bundle in succeeding reflections would be negligible. The history is
then terminated. Such a procedure leads to better accuracy for many
problems because a bundle history extends on the average through many
more events, and a given number of bundles provides a larger number
of events for compiling averages. Haji-Shiekh and Sparrow (ref. 5) have
suggested some other shortcuts for reducing the programming difficulties
of problems involving spectral and directional properties. The obvious
rule of thumb is to use whatever shortcuts can be applied to the case in
question and not be bound by cookbook rules.

EXAMPLE6-1: A gray gas with constant absorption coefficient a is

contained between infinite parallel black plates. Plate 1 is at tempera-
ture T I , and plate 2 is at temperature Tn=O. The plates are separated
by a distance D. Construct a Monte Carlo flow chart for determining
the energy transfer and the gas temperature distribution.
The emission per unit time and area from surface 1 is

If N energy bundles are to be emitted per unit time, then each one must
carry an amount of energy w given by

The bundles are emitted at cone angles /3 given by the first line of table
6-1 of volume I1 as

sin /3= V"&

where R p is a random number in the range 0 to 1. A typical bundle
will travel a path length 1 after emission. The probability of traveling
a given distance S before absorption in a medium of constant absorption
coefficient a is

because of Bouguer's law, equation (1-12). Using equation (6-4) of

chapter 6 in volume 11, this is put in the form of a cumulative distribution
to obtain

However, because Rr is uniformly distributed between 0 and 1, this rela-

tion may as well be written as

where L = LID and K D = aD.

The dimensionless distance normal to the plate X=.x/D that a bundle
will travel when moving through a path length L is then

- cos p
X = L cos p = ------ In R I

Divide the distance D between the plates into k equal increments of

dimensionless width AX=Ax/D, and number the increments w i t h an
increment number j, where

Then the increment number at which absorption occurs is

j = TRUNC (&)+ I
where TRUNC denotes the operation of truncating the value of X / B X
to its integer. At each absorption, a tally is kept of the increment in which
absorption occurs by increasing a counter Sj in the memory of the com-
puter by one unit. This operation is denoted by

If the bundle is absorbed in a gas element, it is immediately emitted

from the same element to conserve energy in this steady-state problem.
This is done by choosing an angle of emission P from the probability
distribution for emission into all cone angles of a unit sphere surrounding
sin p dp
[sin p dp

Using the cumulative distribution function

~,=r 1- cos p

gives the emission angle in terms of a random number as

The distance from the wall to the next absorption point is then given by

X=X,-- cos p In RI

where X, is the position of the previous absorption.

The process of absorptions and emissions is continued until the energy
bundle reaches a black boundary. This occurs when X 2 1 or X 0,
and a counter S,1 or is then increased by one unit to record the
absorption at the black surface.
A new bundle is emitted, and the process is repeated until all N bundles
have been emitted. The dimensionless net energy flux leaving surface 1
is then found from the total bundles emitted minus those reabsorbed

Read i n input:
N, KD, k
Set counters
Sj=O, l s j l k
S w l ' sw2 = 0
Print out results
Q2 = --sw2

Xo = 0 Set path start I

l<j<k at wall 1.

( Determine Choose

6-1.-Flow chart for Monte Carlo solution of radiant transfer between
parallel black plates.

at surface 1; that is,

The net energy flux arriving at surface 2 (-qz) is given by

The temperature at each gas increment is found from equation (6-4) as

and the formulation is complete.

A flow chart is shown in figure 6-1. Note that, since S,l+S,;?=N,

41 - 1- SWl
-- = - q 2
crT: N aT:

and, as expected, ql =- 92; the only reason for printing out both quanti-
ties is to check on the results.
By noting the linearity with T%f this problem, it is possible to gain
solutions for any combination of surface temperatures by use of this
flow chart (ref. 6). Also, by use of the exchange factor relations of section
5.8.3, solutions can be obtained for any combination of gray surface

Some results obtained by the Monte Carlo method will now be



6.4.1 Infinite Parallel Planes
Because of the wealth of solutions available in the literature for a
gray gas between infinite parallel plates, almost every new method of
solution is tried in this configuration and then compared with the results
of one or more of the analytical approaches typified by references 1
and 4.
The Monte Carlo method is no exception, and in reference 6 the

local gas emissive power and the net energy transfer between diffuse-
gray plates is calculated in a manner quite similar to example 6-1.
Parameters are the plate emissivity E (taken equal for both plates) and
various values of the gas layer optical thickness K D = O D where a is
constant. Two cases are examined, the first being a gas with no internal
energy generation contained between plates at different temperatures.
The second czse is a gas with uniformly distributed energy sources
between plates at equal temperatures. Figure 6-2 indicates the accuracy
that can be obtained by Monte Carlo solutions in such idealized situa-
tions. The calculated energy transfer values have a 99.99 percent prob-
ability of lying within 5 5 percent of the midpoints shown.
In figure 6-3, the emissive power distribution within the gas is shown.
Comparison with the exact solutions of references 1 and 4 is quite good;
however, some trends common to all straightforward Monte Carlo
solutions in gas radiation problems are as follows:
First, the calculated individual points in figure 6-3 reveal increasing
error with decreasing optical thickness. This reflects the smaller fraction
of energy bundles being absorbed in a given volume element as the optical
thickness of the gas decreases. As the number of absorbed bundles
decreases, the expected accuracy of the local emissive power becomes
less, and more error naturally appears in the results. Conversely, as the
optical thickness increases, error becomes less; and in figure 6-3, the
curve of results for an optical thickness of ten is quite smooth.
A second effect mentioned in reference 6 is not evident from figure
6-3, and that is that the computing machine time required for solution
of problems involving large optical thickness, say larger than ten, be-
comes quite large. This is simply because the free path of an energy

becomes very short for large optical thickness; therefore, many ab-
sorptions occur during a typical bundle history.
Two limits are now obvious. For small optical thickness, accuracy
becomes poor; for large optical thickness, computer running time be-
comes excessive. From a practical viewpoint, these are not serious
limitations, as the transparent and diffusion approximations to the exact
analytical formulation become valid in just those regions where the
straightforward Monte Carlo approach begins to fail. In addition, the
range of optical thickness over which a Monte Carlo solution can be
effectively utilized can be extended by a variety of techniques, includ-
ing those with the graphic names of "splitting," "Russian roulette,"

Diffusion m l u t i o n (ref. 8)
--- Exact solution of ref. 4
.8 0 Monte Carlo solution (ref. 6)
N = 10 000



0 1 2 3 4 5 6 7 8 9 1 0
Optical thickness, KD

FIGURE6-2.-Heat transfer between infinite parallel gray plates separated by gray gas.


0 Monte Carlo solu-
A .5 tion (ref. 6)
v .I
Analytical solution (refs. 1and 4)

Relative optical depth, ax/aD

6-3. -Emissive power distribution in gray gas between infinite parallel black plates.

and a large number of specialized schemes for specific solutions. Many

of these involve biasing the path length to increase the number of bundles
absorbed in otherwise weakly absorbing regions.

6.4.2 Infinitely Long Concentric Cylinders

A more difficult problem to treat analytically than infinite parallel


- Monte Carlo solu-

P;" .8
h Diffusion solution (ref. 8)
ep .7

E .6

H2 .5
*- .4

E .3

.n .2

Dimensionless distance across annulus,

( r - q)/(ro - ri)

6-4.-Dimensionless emissive power distribution in gray gas in annulus between
black concentric cylinders of radius ratio ri/r,=O.l.

plates is the determination of the emissive power distribution and local

energy flux in a gray gas within the annulus between concentric cylinders.
The energy equation for determining the emissive power at any radius
involves an integral with the local radius appearing as one of the limits.
When the energy equation is written for each incremental layer of gas,
the resulting set of equations have integrals with limits that are differ-
ent for every equation in the set. This set of integral equations must be
solved simultaneously. The Monte Carlo approach, however, differs
only slightly from that for parallel planes. The only additional compli-
cation is the determination of the bundle position in terms of cylindrical
Some Monte Carlo results for an annular region are shown in figure
6-4 as taken from reference 7. Because of the analytical difficulties of
this case, no exact formulation using integral techniques is available
in the literature. Comparison of the results is therefore made with a
modified diffusion solution (ref. 8). Trends in accuracy similar to those
noted for the infinite plate case are evident.

6.4.3 Radiation Between Adjacent Gray Regions

A Monte Carlo formulation has also been applied to study the inter-
action of radiative energy between two regions, each of which has in-

0 Monte Carlo
r e s u s (re. 9)

Diffusion results (ref. 8)

-8 I I
0 .2 .4 .6 .8 1.0
Position i n medium, rlro

6-5.-Dimensionless emissive power distribution in coaxial gas regions with
differing internal energy sources. Radius ratio ri/r,=0.5; optical thickness, 2 (in each

dividual gray radiative properties and internal energy generation rates

(ref. 9). This is of interest here because it gives another insight into
sources of error due to geometrical effects. Figure 6-5 shows the emis-
sive power distribution in two concentric cylindrical regions of the same
optical thickness but different rates of energy generation. The dimension-
less emissive power shown in figure 6-5 has been adjusted for the wall
slip (section 2.6.2) so that the curves all go to zero at the outer wall. The
emissive power was made dimensionless by dividing by the local internal
energy generation rate times the radius of the inner cylinder. The ver-
tical bars give the 95 percent confidence limits on the Monte Carlo
The emissive power in the gas bears a direct proportionality to the
number of energy bundles absorbed in a given volume element. The
volume elements used in the calculation of the results shown in figure
6-5 are of equal radius and, therefore, of differing volumes. The elements
near the center (R + 0) have the smallest volume and, consequently,
the smallest number of bundle absorptions. This is reflected in the in-
creasing width of the 95 percent confidence limits at these points.
Taniguchi (ref. 10) has applied Monte Carlo to radiative transfer in a
gray gas contained in rectangular parallelepipeds.


The greatest criticism leveled against many methods for treating radia-
tive transfer in gases is the inability to accurately account for the strong
spectral, temperature, and pressure dependence of the radiative absorp-
tion coefficient. Such coefficients can sometimes be computed with
reasonable accuracy by quantum-mechanical methods, but few analyses
have been able to include the effect of all variables in the radiative
transfer. Most treatments are limited to gray gases or use of various
types of mean absorption coefficients.
Monte Carlo is well suited to consideration of property variations
with many variables. It involves very little extra effort to assign wave-
lengths to individual energy bundles and to allow the paths of the bundles
to depend on the local spectral absorption coefficient. The relations
necessary to achieve this are given in table 6-1.
If property variations with temperature are considered, an iterative
solution is usually necessary because the temperature distribution
within the medium is not generally known a priori. Determination of
the path length to absorption becomes more difficult also, because the
absorption coefficient varies with position. By applying the formalism
outlined in section 6.3.2 of volume 11, the path length I is found to be
given by

To evaluate this integral to determine 1 along a fixed line after choosing

a random number R is time consuming but at least feasible. Howell and
Perlmutter (ref. 11) used this approach, reducing the complexity some-
what by considering temperature and wavelength dependent absorption
coefficients for hydrogen in the simple geometry of the gas contained
between infinite parallel plates. They considered energy transfer through
the gas between plates at different temperatures, and also the case of
internal energy generation with a parabolic distribution of source
strength in the gas. To evaluate the path length, equation (6-5) was
approximated by dividing the gas into plane increments of thickness
Ax. The path length through a given increment was then

cos p




xz .1



10-5 10-4
Wavelength, A, cm

6-6.-Spectral absorption coefficient of hydrogen at 1000 atm (from ref. 11).

where p is the angle between the bundle path and the perpendicular
to the plates. Equation (6-5) was then replaced by the computational

and the summation was carried out until a value of the integer p was
reached that satisfied the inequality. This value of p would be related
to the increment number in which absorption occurs. The values of
a i , j were assumed for the first iteration and then recalculated in suc-
cessive iterations on the basis of the newly computed local temperatures.
This procedure was continued until convergence was obtained.
Figure 6-6 shows the property variations used, and figure 6-7 shows
a set of emissive power distributions calculated as outlined. The accuracy
becomes poorer, as evidenced by increased scatter, in the regions of
low temperature because of the decrease with temperature of the ab-
sorption coefficient and, therefore, number of absorptions in the low
temperature regions.
Taniguchi (ref. 12) used an incident mean absorption coefficient to
account for property variations.

Monte Carlo with gas

0 .2 .4 .6 .8 1.0
Position between plates, xlD

FIGURE6-7. -Emissive power distribution in hydrogen between infinite parallel plates

at temperatures T I= 9500 K and T2= 4500 K (from ref. 11).


When radiative transfer in a gas is coupled with conductive or con-

vective energy transfer, solution becomes even more difficult. Combined
energy transfer is occurring by radiation, where fourth power tem-
peratures are governing, and by conduction and/or convection, where
derivatives or differences of temperature to about the first power are
governing. The radiative terms in the energy equation for such prob-
lems take the form of multiple integrals, while the conduction terms
contain second derivatives. Further, the radiative surface properties
which appear may be functions of wavelength, direction, and tempera-
ture. When gases are involved, these variables plus pressure can strongly
affect the local gas radiation physical properties. The complete energy
balance on each element of the system then takes the form of a nonlinear
integrodifferential equation.
In the solution of coupled mode problems (refs. 13 to 16), the con-
vective and conductive terms were treated as lumped energy sources
or sinks, and Monte Carlo was used to evaluate the radiative terms on
the basis of an assumed temperature distribution. With the radiative
terms evaluated and substituted into the original equations, conventional
numerical techniques were applied to the resulting differential equations
and a local temperature distribution was generated. This was used as
a basis for reevaluating the radiative integrals, and,the procedure was
continued until convergence.
An example of the power of the Monte Carlo approach in these prob-
lems is given in reference 16. Here the local temperature distribution
as a function of length and radius, and the axial heat flux distribution
in a conical rocket nozzle were determined under conditions expected
in a gas-core nuclear propulsion system. Variations in physical properties
with local temperature, pressure, and wavelength were examined, albeit
not simultaneously, and coupled radiation and convection were con-
sidered. In qddition, the ability was demonstrated of a layer of optically
thick gas injected along the nozzle wall to attenuate the extreme pre-
dicted radiative fluxes to the wall.


The development of Monte Carlo techniques for radiative transfer
under transient conditions has been done by Fleck (refs. 17 and 18) and
Campbell (ref. 19). The model is essentially that outlined in previous
sections, with the additional proviso that the flight times between events
in the history of each bundle are computed, which adds considerable
complexity to the problem. Energy bundles are followed along their paths,
and their position at some time t is used to determine the distribution

of energy at that time. The bundles of course travel at the speed of

light in the medium.
The review article by Fleck (ref. 17) gives a comprehensive discussion
of applications, including the effects of scattering under transient


Scattering of radiation is easily treated by Monte Carlo for any given
distribution of scattering angles. It is analyzed exactly as absorption
and nonisotropic reemission in a gas volume.
Collins and Wells (ref. 20) have used a modified Monte Carlo neutron
diffusion code and other more specialized Monte Carlo codes to study
the transmission of thermal radiation from a nuclear explosion. They
examined the effects of Rayleigh scattering, and of Mie scattering (see
chapter 8) from particles of a given size distribution. Multiple scattering
within an atmosphere of arbitrarily described density distribution and
effects due to ground and cloud reflections were included. Love et al.
(ref. 21) and Stockham and Love (ref. 22) have studied problems with
combined absorption and scattering by Monte Carlo.


The Monte Carlo approach to radiation in attenuating media has been
outlined. Perhaps a sufficient comment as to the power of the method is
made by referring to example 6-1, or more specifically to figure 6-1.
This figure gives a rather complete diagram of the logic required for
programming the problem of energy transfer through a nonisothermal
gray gas between infinite parallel black plates at different temperatures.
A comparison of this diagram with the analyses of, say, references 1 to
4 or chapter 3 will show the simplifications in both concept and formula-
tion that may be present in the Monte Carlo method.

1. USISKIN, C. M.; AND SPARROW, E. M.: Thermal Radiation Between Parallel Plates
Separated by an Absorbing-Emitting Nonisothermal Gas. Int. J. Heat Mass Trans-
fer, vol. 1, no. 1, 1960, pp. 28-36.
2. HOTTEL,H. C.; AND COHEN, E. S.: Radiant Heat Exchange in a Gas-Filled Enclosure:
Allowance for Nonuniformity of Gas Temperature. AIChE J., vol. 4, no. 1, Mar.
1958, pp. 3-14.
3. VISKANTA,R.; AND GROSH, R. J.: Recent Advances in Radiant Heat Transfer. Appl.
Mech. Rev., vol. 17, no. 2, Feb. 1964, pp. 91-100.
4. HEASLET, MAX A.; AND WARMING, ROBERTF.: Radiative Transport and Wall Tem-
perature Slip in an Absorbing Planar Medium. Int. J. Heat Mass Transfer, vol. 8,
no. 7, July 1965, pp. 979-994.

5. HAJI-SHEIKH,A.; AND SPARROW, E. M.: I'rohability Distributions and Error Estimates

for Monte Carlo Solutions of Radiation Problems. Progress in Heat and Mass Trans-
fer, Vol. 2, Pergamon Press, 1969, pp. 1-12.
6. HOWELL,J. R.; AND PERLMUTTER, M.: Monte Carlo Solution of Thermal Transfer
. Through Radiant Media Between Gray Walls. J. Heat Transfer, vol. 86, no. 1, Feb.
1964, pp. 116-122.
7. PERLMUTTER, M.; A N D HOWELL,J. R.: Radiant Transfer Through a Gray Gas Between
Concentric Cylinders Using Monte Carlo. J. Heat Transfer, vol. 86, no. 2, May 1964,
pp. 169-179.
8. DEISSLER,R. G.: Diffusion Approximation for Thermal Radiation in Gases with Jump
Boundary Condition. J. Heat Transfer, vol. 86, no. 2, May 1964, pp. 240-246.
9. HOWELL,JOHN R.: Radiative Interactions Between Ahsorbing Emitting and Flowing
Media with Internal Energy Generation. Paper 66-434, AIAA, June 1966. (See also
NASA TN D-3614, 1966.)
10. TANIGUCHI, HIROSHI: The Radiative Heat Transfer of Gas in a Three Dimensional
System Calculated by Monte Carlo Method. Bull. JSME, vol. 12, no. 49, Feh. 1969,
pp. 67-78.
11. HOWELL,JOHNR.; AND PERLMUTTER, MORRIS:Monte Carlo Solution of Radiant Heat
Transfer in a Nongrey Nonisothermal Gas with Temperature Dependent Properties.
AIChE J., vol. 10, no. 4, July 1964, pp. 562-567.
12. TANIGUCHI, HIROSHI: Temperature Distributions of Radiant Gas Calculated by Monte
Carlo Method. Bull. JSME, vol. 10, no. 42, Dec. 1967, pp. 975-988.
of Rocket Nozzles Using Very High Temperature Propellants. AIAA J., vol. 3, no. 4,
Apr. 1965, pp. 669-673.
14. HOWELL,JOHNR.; AND STRITE,MARYKERN:Heat Transfer in Rocket Nozzles Using
High-Temperature Hydrogen Propellant with Real Property Variations. J. Space-
craft Rockets, vol. 3, no. 7, July 1966, pp. 1063-1068.
15. HOWELL,JOHNR.; AND RENKEL,HAROLDE.: Analysis of the Effect of a Seeded Pro-
pellant Layer on Thermal Radiation in the Nozzle of a Gaseous-Core Nuclear Pro-
pulsion System. NASA T N D-3119, 1965.
Effects in Rocket Nozzles Operating with Very High-Temperature Hydrogen.
NASA TR R-220, 1965.
17. FLECK, JOSEPH A., JR.: The Calculation of Nonlinear Radiation Transport by a
Monte Carlo Method. Statistical Physics. Vol. 1 of Methods in Computational
Physics. Berni Alder, Sidney Fernbach and Manuel Rotenberg, eds., Academic
Press, 1963, pp. 43-65.
18. FLECK,JOSEPH A., JR.: The Calculation of Nonlinear Radiation Transport by a Monte
Carlo Method. Rep. UCRL-6698 (Del.), Lawrence Radiation Lab., Nov. 13, 1961.
19. CAMPBELL,PHILIP M.; AND NELSON,ROBERTG.: Numerical Methods for Nonlinear
Radiation Transport Calculations. Rep. UCRL-7838, Lawrence Radiation Lab.,
Sept. 29, 1964.
20. COLLINS,DAVIDG.; AND WELLS,MICHAELB.: Monte Carlo Codes for Study of Light
Transport in the Atmosphere. Volume 1: Description of Codes. Rep. RRA-T54-1,
Radiation Research Associates, Inc. (ECOM-00240-F, Vol. 1, DDC No. AD-
625115), Aug. 1965.
YIH W.: Radiative Heat Transfer in Absorbing, Emitting and Scattering Medi?.
Oklahoma Univ. (ARL-67-0210, DDC No. AD-666427), Dec. 1967.
22. STOCKHAM, LEO W.; AND LOVE,TOM J.: Radiative Heat Transfer from a Cylindrical
Cloud of Particles. AIAA J., vol. 6, no. 10, Oct. 1968, pp. 1935-1940.
Chapter 7. Energy Transfer by Radiation Combined
with Conduction and/or Convection
When appreciable heat conduction and/or convection occur simul-
taneously with radiation in an absorbing-emitting medium, there are
additional mathematical complications to those already discussed for
radiation alone. Unless it can be shown that both the conduction and
convection have a negligible effect compared with the radiation or vice
versa, a nonlinear integrodifferential equation will result for the energy
equation of the general problem having combined modes of heat transfer.
Fortunately the formulation can be simplified for some circumstances
when all modes must be included. For example, when the gas is optically
thick the diffusion approximation can be applied. The radiation integrals
are replaced by differential terms and a nonlinear differential equation
will then result. Other approximations such as the transparent gas
approximation (section 3.3.1) can be applied under suitable conditions
to simplify the radiative terms.
Because the combined mode problems treated in this Ehapter are
generally mathematically complex, it is not usually possible to obtain
an analytical solution even for seemingly simple physical cases. Con-
sequently, for each physical situation discussed here the analysis will
be formulated and some of the intermediate steps in the solution out-
lined; then the results of a numerical solution are given and discussed.
T h e situations considered will be a stationary conducting and radiating
gas layer between two parallel planes, a boundary layer flow, and a
channel flow of radiating and heat conducting gas.

absorption coefficient
heat capacity
distance between parallel planes; tube diameter
emissive power
exchange factor
function of 7 in Blasius boundary layer solution: gas-to-gas ex-
change factor
surface-to-gas exchange factor

I dimensionless intensity, i/4aT4

i radiation intensity
K, function defined in eq. (7-27)
k thermal conductivity
1 length
Nj conduction-radiation parameter based on temperature, Tj
Pr Prandtl number
energy per unit time
energy flux, energy per unit area and time
heat generation per unit volume and time
radial coordinate
position vector
surface area
absolute temperature
velocities in x-, y-directions
mean velocity
rectangular coordinates
thermal diffusivity
cone angle, angle from normal of area
boundary layer thickness
E emissivity
r) Blasius similarity variable
O dimensionless temperature, TITI
optical depth
cos p
kinematic viscosity
density of fluid
Stefan-Boltzmann constant
dimensionless temperature group (T4- T i ) /(T: - T ; )
slip coefficient defined by eq. (7-25); boundary layer stream
solid angle

b blackbody
c conduction
D evaluated at x = D
e emitted
i incident; inlet

mean value
Planck mean value
Rosseland mean value
evaluated at wall
spectrally dependent
free stream value
surface 1 or 2
in direction of positive cos p
in direction of negative cos p
directional quantity
* dummy variable of integration


There are a number of important practical situations where heat is
transported within a medium by only the two modes of radiation and con-
duction. These usually involve solid or highly viscous media so that the
movement of the medium and hence the convection is not important. In
a liquid or gas, forced and/or free convection are usually of sufficient
importance that they must be included. To develop the theory gradually,
only the radiation and conduction terms are considered in this section;
convection will be added later. The following are three practical situa-
tions where combined radiation-conduction transport is important.
One of the applications is in the glass industry. Although glass is
often thought of as transparent, it can absorb a significant amount of
radiation in certain wavelength regions (see fig. 5-27, vol. I). The
absorbed radiation is then re-emitted within the glass thereby provid-
ing a radiative transport traveling layer by layer through the medium.
The ordinary thermal conduction is thus augmented by a "radiative
conduction." Radiative effects are quite important in influencing the
temperature distribution within molten glass in a furnace. These ef-
fects have been analyzed by Kellett (ref. l), Gardon (ref. 2), and Condon
(ref. 3).
A second application is concerned with glassy materials that are some-
times used as an ablating coating to protect the interior of a body from
high external temperatures by sacrificing the ablating surface. The

radiation-conduction process is important in regulating the temperature

distribution within the ablating layer. The temperature distribution in-
fluences how the ablating material will soften, melt, or vaporize. These
processes determine how efliciently the material will protect the surface.
The radiative analysis of an ablating material has been considered by
Kadanoff (ref. 4).
A third area of application that has arisen more recently is radiation
within cryodeposits of solidified gas that form on a very cold surface.
The cold surface may be on a space vehicle orbiting at the upper fringe
of the atmosphere, or may be part of a cryopump which is a device used
to produce a high vacuum by condensing the gas within a chamber. The
cryodeposit coating changes the radiative properties of the cold surface
and can thus significantly influence the radiative exchange with this
surface. The radiative transfer in cryodeposits is considered in references
5 and 6.
In this section, some methods are examined for treating energy
transfer by combined radiation and conduction. The conduction-radiation
parameter will be introduced and the energy equation formulated. Then
some approximations are considered, the most simple being the addition
of separately computed radiation and conduction energy transfers to
obtain the combined transfer. Any of the approximation methods to the
equation of transfer presented in chapter 3 can be applied to simplify
the radiation terms in these multimode problems, and the diffusion
method will be applied here as an example. Also, the application in the
literature of the Monte Carlo technique for combined mode problems
has been mentioned in section 6.6.

7.3.1 The Conduction-Radiation Parameter

When conduction is present, a new dimensionless conduction-radiation

L Medium with thermal conductivity k

and absorption coefficient a

7-1. -Conduction through and radiation from plane volume element.

parameter N is introduced. The definition of N can be developed by

analyzing the one-dimensional layer of material shown in figure 7-1.
The material has thermal conductivity k, absorption coefficient a , and
is one radiation mean free path thick. If it is assumed that the tempera-
ture profile within the slab is nearly linear, the conduction through the
layer for area A is

The total radiation emitted by the layer of area A can be written by using
equation (1-34) as (not including induced emission)

by suitably defining a mean temperature T,,, and neglecting attenuation

in the volume. The ratio of conducted to emitted flux is then

Dividing through by Ti and letting O = TITI give

The Nj is the conduction-radiation parameter based on the jthtempera-

ture, defined by

For the special case when Of, = 1 - 02 then QJQe = N 1 , and the
parameter gives a measure of the relative energy amounts carried by
conduction and emitted radiation for this layer of thickness lla.
Generally, however, Nj does not directly give the relative values of
conduction to emission because the ratio of these values depends also
upon both the temperature difference and temperature level in addition
to N j as shown by equation (7-4).

7.3.2 Energy Balance

To obtain an analytical solution for the combined conduction-radiation
energy transfer in an absorbing-emitting medium, the general energy
equation must be formulated. This is then solved subject to the boundary
conditions to obtain the temperature distribution in the medium, and the
heat flow can then be found. To fornlulate the energy equation, equation
(2-16) is used for the total energy absorbed by a volume dV from the
incident radiation intensity, and equation (2-17) gives the total energy
emitted. In addition there will be a net energy gain per unit volume by
heat conduction equal to kV2T for the case where k is a constant.
Equating the net gain by conduction to the net loss by radiation gives

If there is internal heat generation in the medium, t h e heat thereby

added per unit volume and time must be added to the left side of equation
(7-5). The heat generation per unit time and volume will be designated
as 9"' and can be a function of position and time in t h e medium. This
generation can be produced for example by electrical, chemical, or
nuclear means.
For a transient problem some of the heat inflow to the volume element
can be stored in the element. The energy storage per unit time and
volume is p c l , ( a T / a ~ )Then
. for a transient situation with internal heat
generation the energy equation becomes

The transient case for pure radiation problems is treated in section 9.6.
Since the radiation terms in equations (7-5) and (7-6) depend not
only on the local temperature but on the entire surrounding radiation
field, the energy equation is an integrodifferential equation for the tem-
perature distribution in the medium. The conduction and heat storage
terms depend on a different power of the temperature t h a n the radiation
terms and the energy equation is thus nonlinear.
Numerical solutions of the energy equation have b e e n carried out by
Gardon (see section 9.3.2), Viskanta and Grosh (refs. 7 a n d 8),and others.
Most solutions have been for media in a plane-slab geometry, although
some solutions have been carried out for other configurations (refs. 9 and

By inserting equations (7-9a) and (7-9b) into equation (7-8), the i' is
eliminated and an energy equation for T(x) is obtained which can be
solved numerically. However, some further simplifications in form are
Define the nondimensional quantities p = cos P , 0 = TITl, 0 2 = T ~ / T I ,
N1 = ka/4aTf, K = a x , KII = aD, I' = ir/4aT?. Then by also using
d o = 2~ sin @dB= - 27rdp, equations (7-8) and (7-9) can be placed
in the form

d 2 0 ( ~-)
NI -
7 - @*(K)- 271.
I,' I ; ( K ,p ) d p - 271. I,' I: (K, p ) d p (7-10)


Note that the signs have been changed by changing variables so that p
is positive throughout equations (7-10) and (7-11). Combining equations
(7-10) and (7-11) to eliminate I' gives

By using the exponential integral function defined in equation (2-4.5)

and in the appendix, equation (7-12) can be written as

Equation (7-13) is the desired integrodifferential equation for the tem-

perature distribution @ ( K ). It is a nonlinear equation since O is raised to
the first power in the conduction term while it is raised to the fourth
power in the radiation terms. The boundary conditions in dimensionless
form are
@=I at K = O
@=@% ~ ~ K = K D

Examining equations (7-13) and (7-14) shows that the solution depends
on the parameters N1, KD, and On.
In addition to the temperature distribution, the heat transfer across
the layer from plate 1to plate 2 is usually of interest. Equation (2-41) gave
the net heat flux expression for radiation alone across a gray gas between
black plates. This radiative flux equation was obtained for convenience
at x = O . In addition at the same location there is now a conduction flux
-k(dT/dx) I x = o so that the heat flux equation becomes (note that from
energy conservation q will not depend on x for the situation being con-
sidered here)

q = - k d r dx
l + U T ! - ~ % ~sin

On the right, the first term is the conduction away from wall 1, the sec-
ond is the radiation leaving wall 1, the third is the radiation leaving wall
2 hat is then attenuated by the medium and reaches wall 1 , and the
last term is the radiation from the medium to wall 1. By using the expo-
nential integral function and the previously defined dimensionless vari-
ables, the heat flux can be written as


0 .2 .4 .6 .8 1.0
Relative optical depth, K I K ~

FIGURE 7-2. - Dilnensionless temperature distribution in gray gas between infinite parallel
black plates with conduction and radiation. Plate temperature ratio O e = O . l ; optical
spacing K O = 1.0. (Data from ref. 7.)

Since there are no heat sources in the medium, this q computed at the
lower wall is the same at all locations within the medium.
Viskanta and Grosh (refs. 7 and 8) obtained solutions of equation (7-13)
by numerical integration and iteration. Some of their temperature
distributions are shown in figure 7-2. For N 1 -+ co conduction dominates
and the solution reduces to the linear profile for conduction through a
plane layer. When N1= 0, the conduction term drops out and the
temperature profile has a discontinuity (temperature slip) a t e a c h wall
as discussed for the case of radiation alone in section 2.6.2. When
conduction is present, there is no temperature slip. Some of the heat
flux results from reference 8 as obtained from equation (7-16) a r e shown
in table 7-1.
Timmons and Mingle (ref. 11) have carried out solutions for t h e same
problem with specular, rather than diffuse, boundaries. Results are
within a few percent of those for diffuse boundaries.

[I'rrm ref 81

Optical Plate Conduction- Dimensionless

thickness, temperature radiation energy flux,
KD ratio, Or parameter, N1 q/u'l';

0.1 0.5 0 0.859

.01 1.074
.1 2.880
1 20.88
10 200.88

1.0 0.5 0 0.518

.01 .596
.1 .798
1 2.600
10 20.60

1.0 0.1 0 0.556

.01 ,658
.1 ,991
1 4.218
10 36.60

10 0.5 0 0.102
.01 .114
.1 ,131
1 ,315
10 2.114

7.3.4 Simple Addition of Radiation and Conduction Energy Transfers

A relatively simple idea to obtain the combined energy transfer by
radiation and conduction is to assume that the interaction between the
two transfer processes is so small that the processes can be considered
to each act independently. Then the conduction and the radiation
transfers are each formulated as if the other transfer mechanism were
not present. Einsteinyref. 12) and Cess (ref. 13) have investigated this
approximation for an absorbing-emitting gray medium between infinite
parallel plates. When the plates were black, the results for the energy
transfer were within 10 percent of the exact solution. Larger errors are

Tltis is T h n m a s H. Einstein. Albert made his appearance i n chapter 4


possible if highly reflecting surfaces are present. Howell (ref. 14) shows
that the additive solution is also fairly accurate for a gray gas between
black concentric cylinders.
An additive solution cannot be used to predict teillperature profiles.
It is an effective and simple method of predicting energy transfer by
combined modes, although the accuracy of the solutions so obtained
becomes doubtful in some situations. The use of the additive method
for problems where the accuracy has not been established by some
comparisons with more exact solutions is not advisable.

EXAMPLE7-1: By using the additive approximation, obtain a relation

for the energy transfer from a gray infinite plate at temperature TI and
emissivity € 1 , to a parallel infinite gray plate at Tz with emissivity €2.
The spacing between the plates is D, and the region between the plates
is filled with a gray material with absorption coefficient a and thermal
conductivity k. Use the diffusion approximation for the radiative transfer.
Without radiation, the energy flux from surface 1 to 2 by pure conduc-
tion to be used in the additive solution is

The diffusion solution for pure radiation from plate 1 to 2 can be found
from table 3-11 as

Since the two modes are assumed completely independent, the additive
solution gives
4 = q c + 9,. (7-1 9)

After using the dimensionless variables defined in connection with

equation (7-lo), the expression for q becomes

Equation (7-20) must give correct results at N1 = 0 (pure radiation)

within the accuracy of the diffusion solution and at N1 -+ (pure con-

duction) because it simply adds these two limiting cases.

Comparison of q/o-7': from equation (7-20) with exact numerical

solutions for E l = €2 = 1 and = 0.5 from the work of Viskanta and
Grosh (refs. 7 and 8) is shown in figure 7-3. For this geometry and
black surfaces, the results of the additive solution are very accurate.
The additive method appears even better here because of the fortuitous
benefit that the diffusion solution gives a pure radiation heat transfer
that is slightly above the exact pure radiation solution (fig. 3-6) while
the pure conduction result is too low. This is because the conduction
solution is based on the linear gradient of T at the boundaries while the
actual gradients are larger whetl radiation is present (see fig. 7-7(a) for
example). The errors in the two solutions tend to cancel, thus giving a
quite accurate combitled solution for this geometry.


0 Exact numerical solution (refs. 7 and 8)
Additive solution (eq. (7-20))

Conduction-radiation parameter, N1

7-3. -Col~lparison of sinlple additive arid exact riumerical solutiolls of combined
conduction-radiation energy transfer between black parallel plates. Plate temperature
ratio 0,=0.5.

7.3.5 The Diffusion Method

This ~netllodhas the advantage over the additive method in that a
solution is obtained to the coupled energy equation and this yields the
temperature distribution in the medium. In section, it was
shown that the diffusion heat flux for radiative transfer has the same
form as the Fourier conduction law. By using the ~ o s s e l a n dmean
absorption coefficient defined in equation (3-39), the radiative flux vector
can be written as

where k, is the radiative conductivity defined by

Consequently, by use of the diffusion approximation, the energy flux

vector by combined radiation and conduction at any in the
medium can be expressed as

The local heat flux as given in equation (7-23) can be u s e d as in

the derivation of the heat conduction equation, to obtain a n energy
balance on a differential volume element within the absorbing-emitting
medium. For example, in two-dimensional rectangular coordinates and
with no internal heat sources, the energy equation is

The medium behaves like a conductor that has a thermal conductivity

dependent on temperature.
To obtaln the temperature distribution in the medium, a n equation
such as equation (7-24) must be integrated subject to t h e imposed
boundary conditions. These conditions would often be the t e m p e r a t u r e s
on the enclosure surfaces. However, near a boundary t h e diffusion
approximation is not valid; consequently, the solution is incorrect
near the wall and it cannot be matched directly to the wall boundary
conditions. To overcome this difficulty, the boundary condit30n at the
edge of the absorbing-emitting medium will be modified S n such a
way that the resulting solution to the diffusion equation will b e correct
in the region away from the boundaries where the diffusion aPproxi1nation
is valid.

I n the pure radiation c a s e , a temperature slip was introduced to over-

come t h e difficulty of matclling the diffusion solution in t h e medium to
the wall temperature. For combined conduction-radiatioll problrms, a
similar slip concept was introduced hy Goldstein a n d Howell (rrfs. 15 and
16). Ey using t h e method of matched asymptotic expansions. which in
this c a s e was used to match the linearized solutions for intensity,
flux, and temperature near t h e wall with t h e diffusion solution for
these quantities far from the wall, a n effective slip condition was de-
lived. As shown by figure 7-4, this slip gives t h e boundary conditio~l
T(n+O) that the diffusion solution must have if it i s to extend all t h e way
to t h e wall. T h e slip is given in terms of t h e slip coeficient $ which is a
function only of the conduction-radiation parameter N. I11 terms of
quantities at wall 1, is given by

where q , is the radiative energy flux at the boundary as evaluated by

t h e diffusion approximation, TI is the wall temperature, and T ( x + 0) is
the extrapolated temperature i11 t h e medium at t h e wall ( t h e effective
slip temperature to b e used in the diffusioil solution). T h e $1 is com-

?Exact solution for ternper-

,/ ature profile i n gas

FIGURE7-4. -Use of effective temperature slip as boundary condition for diffusion solution
in combined conduction and radiation.

FIGURE7-5. -Slip coefficient for combined conduction-radiation solutions by the diffusion


puted from the relations of reference 15 as

where I<, is the function

A graph of $ as a function of N that can be used for any geometry is shown

in figure 7-5.
This type of solution should give accurate temperature distributions
within the limits of the assumptions inherent in the diffusion approxima-
tion. Results can be obtained for both the energy transfer and the tem-
perature profiles as will be shown by an exainple problem. Other solu-
tions of this general type have been presented in references 17 and 18.

EXAMPLE7-2: Using the diffusion method for combined conduction

and radiation, find an equation for the temperature profile in a medium
of constant absorption coefficient a and thermal conductivity k . The
medium is contained between infinite parallel black plates at temper-
atures TI and TZ spaced D apart with the lower plate 1 located at x= 0.

Equation (7-23) in dimensionless form for this geometry becomes the

following (note that an= a in this case):

From energy conservation, since there are no internal heat sources

the q is constant across the space between plates. Equation (7-28) can
then be integrated from 0 to KD to yield

where @ ( 0 )and O ( K ~are

) in th,e medium at the lower and upper bound-
aries, respectively. These two temperatures must now be eliminated by
using the slip boundary conditions to relate them to the specified wall
temperatures T1 and Tz.
Consider first the boundary condition at wall 1. For the particular N1
of the problem, the $1 is found from figure 7-5 and set equal to

From equation (7-23) the radiative flux at the wall qr,l can be written a s

Then $1 becomes

u[T: - T 4 ( 0 ) ]
$1 =

This is rearranged into


As shown in derivation of $ (ref. 15), the conditions where the diffusion

solution is valid lead to the jump T1 - T(0) being small. For convenience
a portion of equation (7-30) can then be linearized. With T1 - T ( 0 ) = 6
where 6 is small, then

Then equation (7-30) becomes

or in dimensionless form

Similarly at wall 2 (note that the value of $2 corresponds to N = N2 on

fig. 7-5)

Now add equations (7-29), (7-31), and (7-32) to eliminate the unknown
temperatures in the medium 0 ( 0 ) and O ( K ~ )This
. yields the nondi-
mensional energy flux transferred across the layer as

The energy flux results of equation (7-33) are plotted in figure 7-6
and compared with the exact and additive solutions (the exchange factor
approximation shown in the figure will be discussed in the next section).
At K U = ~the
, results compare very well with the exact solution. For a
small optical thickness K ~ ~ 0 . however,
1, the diffusion-slip procedure
breaks down for intermediate values of N1 and the simple additive solu-
tion provides much better energy transfer values.
An advantage of the diffusion solution is that it will yield the tem-
perature distribution in the medium. Temperature profiles can be pre-
dicted by integrating equation (7-28) from locations 0 t o K and then

Solution Source

I- /,/
0 Exact numerical
A Exchange factor approximation Ref. 14
Simple additive Eq. (7-20)
----- Linearized diffusion slip Ref. 15 .


0. 1

Conduction-radiation parameter, N1

7-6. -Comparison of various methods for predicting energy transfer by conduction
and radiation across a layer between parallel black plates. Plate temperature ratio
T2/T1= Oe=0.5.

using equations (7-31) and (7-33) to eliminate 0(0) and q. This yields

Some temperature profiles are shown in figure 7-7. For K I I = ~ , figure

7-7(a), the profiles are poor except for the largest N1 shown. Better re-
sults are obtained for all N1 at larger KD because the assumptions in the
diffusion solution become more valid. This is shown by figure 7-7(b),
where results are compared for K g = 10. On the basis of the assumptions
used in the diffusion slip analysis and the way in which the solutions
compare with exact analytical solutions, good temperature distribution

- Exact numerical solution

--- Diffusion slip approximation

Relative optical depth, K/K~

(a) Optical thickness KD= 1. (b) Optical thickness Ku=lO;

conduction-radiation param.
eter Nl = 0.02916.
7-7.-Comparison of temperature profile by exact solution (ref. 7) with diffusion
slip approximation. 131ate temperature ratio Q2=0.5; plate ernissivities E I = E Z = 1.0.

results are expected for K g 2. For N1+ 0 and N1-+ w, the diffusion slip
method goes to the correct limiting solutions.
Within their limits of applicability, diffusion methods provide a dif-
ferent interpretation of the conduction-radiation parameter from that
presented in section 7.3.1. The ratio of the molecular conductivity k to
the radiative conductivity k, given by equation (7-22) is

Therefore, in the diffusion limit, N is a direct measure of the ratio of

molecular to radiative conductivity. In consequence, N is in this case also
a direct measure of the ratio of the energy transferred by the two modes.

7.3.6 The Exchange Factor Approximation

In section 5.8.3 the use of the exchange factor approximation was
introduced in the context of pure radiation problems. In this section,
these exchange factors are applied to combined mode problems, and
it is shown that their use gives a convenient method of treating these
problems. As stated in section 5.8.3, the exchange factor F j - k is defined
as the fraction of the energy leaving surface j that is incident on surface

h- when all boundaries are black and the intervening medium is in radia-
tive equilibrium (i.e., when radiation is the only means of energy ex-
change). The F i - k includes the effect of absorption and re-emission of
energy by the medium while the energy is in transit from A, to Ak. A
similar exchange factor between a surface area and volume element was
also defined there. The reader may find it helpful to review section 5.8.3
before proceeding with the present material.
The total energy emitted by a volume element in the presence of heat
conduction can be equated to three terms: (1)the energy that would be
emitted in the absence of conduction, (2) the net energy supplied to the
element by conduction (and thus must be radiated away), and (3) any
additional energy dQ,,t,., added to the element by radiation over and
above that given by the radiative equilibrium (zero conduction) case
because of the change in temperature profile as a result of conduction
being included. This can be written as

The temperature TI? present in the radiative equilibrium case is con-

veniently written in terms of exchange factors as in equation (5-122) as

The Q,, j is the energy leaving the jth

surface of the enclosure surrounding
the gas. The clFj-( the fraction of energy leaving the area Aj that is
absorbed in clV for the case of radiative equilibrium in a black enclosure
and includes the effects of gas absorption and emission while the energy
is in transit from A j to dV. Substituting this relation and assuming that
dQ,,t,., is small reduce equation (7-36a) to

Note that the exchange factors include the effect of gas-to-gas volume
element radiant interchange based on the temperature profile for no
conduction (radiative equilibrium). The approximation introduced
in equation (7-36b) is that the gas-to-gas radiant exchange is not signifi-
cantly affected by the new temperature profile that results because of
the presence of gas conduction (i.e., dQeXt,.(, 0). The similar approxi-

mation is also made that the Q , , j can be used from the solution without
conduction. If radiation predominates, then this is a good assumption;
if radiation is small, then it will not matter that the radiative terms are
somewhat inaccurate.

Equation (7-36b) is a nonlinear differential equation for T, the local

gas temperature. Howell (ref. 14) has applied this approach to gray gases
in annular enclosures and between infinite parallel plates. Accuracy is
found to be comparable to the simple additive solution when computing
the heat flux through the gas as shown in figure 7-6. The chief advantage
of the method is that accurate temperature distributions can be obtained
for combined-mode energy transfer problems with but little effort. The
procedure will be demonstrated by the following example.

EXAMPLE7-3: Find an expression that will yield the temperature

profile in a gray gas contained between infinite parallel plates a distance
D apart if the gas has absorption coefficient a, thermal conductivity k,
and the plates are black at temperatures TI and T b Use the exchange
factor approximation.

Using a layer dx thick as the volume element results in governing

energy equation (7-36b) becoming, in this geometry,

Exact numerical solution

----- Exchange factor approximation
1.0 Conduction -





0 .2 .4 .6 .8 1.0
Relative optical depth, K / K ~

7-8.-Co~nparison of temperature profiles by exact solution with exchange factor
approximation. Optical thickness K n = l . O ; plate temperature ratio @ e = O . l ; plate
emissivities e l = ey= 1.0.

The exchange factors are given by equations (5-125) and (5-126) as

d F l -d, = 4(1dx PI,(x)
and -
dFe-d.T=4ad~(l - p b )

The cpb can be obtained to good accuracy for this geometry by the dif-
fusion relations, table 3-11, as

Substituting these relations into equation (7-37) and using the usual
nondimensional quantities give

which is to be solved subject to the boundary conditions 0 = 1 at K = O

and O = 0 2 at K = KD. This relation gives the correct limiting diffusion
solution for N 1 + O and the correct conduction solution for N l + w .
Howell (ref. 14) has solved for [O ( K )- O2]/ (1- 02) numerically, using
exchange factors from the numerical solutions of the pure radiation
problem which are more accurate than the diffusion exchange factors
used in equation (7-38). Agreement with the numerical solution of the
coupled conduction-radiation problem is shown in figure 7-8.

The energy transfer by conduction was found by numerically evaluating

d @ / d ~ I . =and
~ using this to evaluate the conduction flux at the boundary.
The radiation flux was assumed to be unaffected by the conduction process
in the spirit of the exchange factor approximation. The results obtained
in this approximate way agree quite well with the numerical solution
as shown in figure 7-6. Since d @ / d varies
~ with K, while the radiative
flux without conduction is constant with K, evaluating d @ / d a~t a location
other than K = O would give different results. For the most accurate

calculation of heat flux it was found (ref. 14) that the tenlperature gradi-
ent should be evaluated at the boundary with the highest temperature.
Using the gradient at the coldest wall leads to a heat flux prediction that
is always too large.

The exchange factor approximation is not limited to any range of

geometry, surface emissivity, or optical thickness; it can be applied to
any situation where the exchange factors have been previously obtained
or where they can be obtained by some simplified pure radiation solution.
The resulting nonlinear energy equation can usually be cast in the form
of a matrix of nonlinear difference equations which are often easily
solved by the numerical technique outlined by Ness (ref. 19). (See section
7.3.2, in vol. TI.)
In situations where the boundaries are not the chief contributors to
the radiant energy flux in the gas, then the gas temperature profile
becomes important to the radiative flux distribution. In such a case the
exchange factor approximation may become inaccurate; however,
reference 14 does give results that compare well with exact numerical
solutions for both energy transfer and temperature profiles in the geome-
tries of parallel plates and concentric cylinders.
Lick (ref. 20) has also presented various approximations for solving
conduction-radiation problems. He develops methods for treating gases
with spectral and temperature-dependent properties. Goldstein and
Howell (ref. 15) outline methods of treating temperature-dependent
properties using the apparent slip technique.
Although numerical methods are the only way to obtain exact solutions
to combined mode problems, the approximate methods outlined in this
chapter should provide acceptable accuracy for most engineering prob-
lems. All the methods presented here can be applied to problems in
two and three dimensions.


The interaction of convection, conduction, and radiation in absorbing-
emitting media occurs in many practical cases. Atmospheric phenomena,
shock problems, rocket nozzles, industrial furnaces- all these and many
more involve such interactions. As a consequence, a large amount of
literature is available. Review articles and comprehensive books are
given by references 13 and 21 to 25. In spite of this material, these prob-
lems remain difficult to solve. In this section, some of the methods used
for these problems are outlined.

7.4.1 Boundary Layer Problems

A number of papers have appeared that deal with the effect of radia-
tion on forced convection boundary layer heat transfer (refs. 26 to 30).
Such boundary layer problems are approached by writing the usual con-
tinuity and momentum equations: these do not contain any radiation
terms and are not influenced by heat transfer since constant fluid prop-
erties are assumed here. The energy equation is then written including an
energy source term to account for the net thermal radiation gained by
a volume element.
For a two-dimensional boundary layer flowing over a flat plate (fig.
7-9), assuming a gray fluid with constant properties, negligible viscous
dissipation, and negligible radiative transport in the flow direction,
the energy equation is

where q, is the radiation flux in the positive y-direction shown in figure

7-9. The problem becomes one of introducing into equation (7-39)
one of the formulations for q, that have been developed, and then solving
the resulting energy equation together with the momentum and continuity
equations. In references 27 and 28, the diffusion solution is used for q ,
and the last two terms in equation (7-39) can then be combined as was
done in equation (7-23).
Various techniques can be used to solve the resulting energy equation.
Novotny and Yang (ref. 26) used matched asymptotic expansions of the
energy equation assuming a known flow field. A linearized energy equa-
tion near the surface was matched to an asymptotic solution far from the
surface. Viskanta and Grosh (ref. 27) had earlier applied the diffusion
approach; they assumed that the diffusion solution was valid all the way
to the boundary.
Cess (ref. 28) and others have assumed the boundary layer to be
optically thin, so that it emits but does not absorb radiation. By introduc-
ing some other assumptions, Cess was able to treat nongray gas effects.
Neglecting absorbed radiation can be a useful approximation when the
boundary layer is heated by frictional dissipation while surface and sur-
rounding gas are cool.
Outer region
-uo, To

LTl (constant)

7--O.-Bou11darylayer flow over flat plate.


Fritsch, Grosh, and Wildin (ref. 29) studied the shielding of a body
by a radiant absorbing layer. The boundary layer was assumed to
absorb, but not to emit radiation. The effects of transpiration and an
external radiation field were included. Howe (ref. 30) had treated a similar
problem under somewhat different conditions of the external radiation
Novotny and Kelleher (ref. 31) and Cess (ref. 32) have examined the
effect of radiation on the boundary layer development in free convection.
An absorbing-emitting gas was treated and the radiation effects were
linearized in both cases. Reference 31 treats the development of the
boundary layer on a horizontal cylinder, while reference 32 examines the
layer growth on a vertical plate. Gille and Goody (ref. 33) experimentally
determined the onset of free convection in a gas exposed to thermal
radiation. Optically thin thermal layer.-Let us now consider in more
detail the analysis for laminar forced flow on a flat plate. An expression
is needed for the radiative source term dq,./dy in equation (7-39). Within
the boundary layer type of assumptions it will be assumed that the
thermal conditions are changing so slowly in the x-direction, as com-
pared with the changes in the y-direction, that the region contributing
to q, at a specific x, say x+, are all at that x+ and hence at temperatures
T ( x f , y). Then qr can be evaluated from the one-dimensional relations
derived previously as for equation (7-13). In equation (7-13) which is
written for a region between two black walls and no convection, the
conduction term is equal to the following (using y as the transverse

k -=4aaT"2aa T:E~(K)-I- T$P(KD- K)

The right side of equation (7-40) is thus the dq,/dy term when two walls
are present. For one bounding wall, the Tp term is not present and the
upper limit of the integral is extended to infinity. Also the T 4 ( ~ * is
replaced by T q x , K*) to emphasize the approximation that, for the
radiation term, the surroundings of position x are taken to be at T(x, y).
Then the boundary layer equation (7-39) becomes for the temperature
T(x, Y)

where K = ay.
The temperature field can be considered as composed of two regions.
Near the wall in the usual thermal boundary layer region of thickness
6 that would be present in the absence of radiation, there are large
temperature gradients and heat conduction is important. This region is
usually of small thickness; hence, it can be assumed optically thin so
that radiation will pass through it without attenuation. For larger y than
this region, the temperature gradients are small and heat conduction is
neglected compared with radiation transfer. The approximate analysis
can now proceed, for example, along the path developed in reference 13.
In the outer region the velocity is the free stream value uo, and with
the neglect of heat conduction the boundary layer equation reduces to

To obtain an approximate solution by iteration, substitute the incoming

free stream temperature To for the temperature on the right side as
a first approximation and then carry out the integral to obtain a second
approximation. This yields for the outer region to first-order terms

where at x= 0, T= TO.
At the edge of the thermal layer ~ = a y = a G which is small so that
E2( a 6 ) = ES(0) = 1. Hence, at y= 6 equation (7-43) becomes

Equation (7-44) is the edge boundary condition that the outer radiation
layer imposes on the inner thermal layer. The outer temperature is
increasing linearly with x. This is the result of the flowing gas ab-
sorbing a net radiation from the plate in proportion to the difference
T: - T.',and the absorption coefficient a.
To solve the boundary layer equation in the inner thermal layer
region, the last integral in equation (7-41) is divided into two parts,
one from K = O to a6 and the second from a6 to m. The first portion is
neglected as the thermal layer is optically thin, and the second is evalu-
ated by using the outer solution (eq. (7-43)). By retaining only first-
order terms, the boundary layer equation reduces to

The boundary collditiolls are given by equation (7-44) at y= 8, and the

specified wall temperature T= TI at y= 0.
The solution becomes quite complex and will not be further developed
here. The reader is referred to references 13 and 34 for additional
information. Optically thick thermal layer. -At the opposite extreme from
the previous section, if the thermal layer has become very thick or the
medium is highly absorbing, then the boundary layer would b e optically
thick. In this instance the analysis is considerably simplified as the
diffusion approximation can be employed. Referring to equation (7-23)
it is recalled that radiative diffusion adds a radiative conductivity to
the ordinary thermal conductivity. Then the boundary layer energy
equation (7-39) can be written as

With the assumption of constant fluid properties, the boundary layer

momentum equation and the continuity equation do not depend on tem-
perature. Consequently, the flow is unchanged by the heat transfer,
and velocity distribution is given by the Blasius solution (ref.
35). The Blasius solution is in terms of a similarity variable

and the stream function and velocity comporients are given by

These quantities are substituted into the energy equation w h i c h can then
be placed in the form

The boundary conditions that have been used are

To be more precise, a diffusion-slip condition should be used at the wall

but this has not been formulated for the combined radiation, convection,
and conduction situation.
Let O = TITO and No= kafi/4aTo3;then equation (7-47) becomes

Numerical solutions were carried out in reference 27, and some typical
temperature profiles are shown in figure 7-10. For No=lO the profile
was found to be within 2 percent of that for conduction and convection
alone (i.e., for No-+ m). The effect of radiation is found to thicken the
thermal boundary layer similar to the effect of decreasing the Prandtl
number. This would be expected since Prandtl number is the ratio of
viscous to thermal diffusion v / a . The radiation has supplied an additional
means for thermal diffusion thereby effectively increasing the a.

FIGURE 7-10.-Boundary layer temperature profiles for larnii~arHow 011 Hat plate (ref. 27).
Prandtl number Pr= 1.0; temperature ratio T I / T o = O . ~ .

7.4.2 Channel Flows

Of engineering interest in some high temperature heat exchange
devices is the flow of an absorbing-emitting gas in a channel with radiative
and convective energy transfer. The energy equation (17-39) (with v = O
for fully developed Aow) still applies at all points in the flow. References
12 and 36 to 39 treat problems with varying degrees of approximation.
Viskanta (ref. 36) gives approximate numerical solutions to the equations
for laminar flow of a gray absorbing-emitting medium in a parallel plate
channel. Temperature-independent properties are assumed. In addition
to K D , N, and temperature ratios, a new parameter enters these problems.
It is a Nusselt number which is defined in such a way a s to include a
radiative contribution and thus differs from the usual definition.
Einstein (refs. 12 and 37) has applied the gas-to-surface and gas-to-gas
view factor methods of Hottel (section 5.8.2) to a solution of the energy
equation in a parallel plate channel and in a circular tube. Both channels
were of finite length, and internal heat generation in the gas was included.
Conlparison was made with the work of Adrianov and Shorin (ref. 38),
who had used the cold material approximatioil (section 3.3.3) so that
absorption but not emission from the gas was included. Chen (ref. 39)
included scattering in his analysis of flow between parallel plates, but
assumed a slug flow velocity profile.
All the previous analyses mentioned in this section have been for
gray gases flowing in channels with gray or black walls. All have as-
sumed temperature-independent properties. However, deSoto and
Edwards (ref. 40) have presented a tube flow heat transfer analysis that
accounts for nongray gases with temperature-dependent properties. An
exponential band model (section 4.6.4) was employed to account for
the spectral effects. Entrance region flows were included.
Landram et al. (ref. 41) examined fully developed turbulent pipe
flow of an optically thin gas, using Planck and incident mean absorption
coefficients. The application of Monte Carlo methods to some channel
flow problems was discussed in chapter 6.
To examine a specific situation, consider the analysis of Einstein
(ref. 37) for flow in a tube of diameter D as shown in figure 7-ll(a).
Gas enters the tube at temperature T iand leaves at To. The tube wall
temperature is constant at T,. The surrounding environments at the
inlet and exit ends of the tube are assumed to be at the inlet and exit
gas temperatures, respectively. The governing energy equation at
position -7 in the tube for laminar flow can be written as

The triple integral is the radiation absorbed at T: as a result of emis-

sion from all the other gas in the tube. The f ('i"-7) is a gas-to-gas
exchange factor from position 7 " to position 7 . The double integral
is the radiation absorbed at' 7 as a result of emission from the bound-
aries which include the tube wall and the end planes of the tube. The

0 2 4 6 8 1 0
Optical thickness, KD = aD

(a) Tube geometry and boundary conditions.

(b) Exit temperature for T i / T , , = 0 . 4 , l / D = 5 , and puc,,/uFh = 3 3 .
7-11.-Combined radiation and convection for absorbing gas flowing in tube with
constant wall temperature (ref. 37).

g.(T:*- T : ) is a surface-to gas exchange factor, and the f and g are given

in reference 37.
Some typical results of the numerical solution are given in figure
7-ll(b) for a Poiseuille flow velocity profile. These results show how
well the gas obtains energy from the wall since the ordinate is a measure
of how close the exit gas approaches to the wall temperature. The results
are given in terms of gas optical thickness based on tube diameter and
a conduction-radiation parameter based on wall temperature. As the
optical thickness is increased from zero, the amount of radiated energy
from the wall that is absorbed by the gas increases to a maximum value.
Then for a large K D the heat absorbed by the gas decreases. The decrease
is caused by the self-shielding of the gas; for high K D most of the direct
radiation from the tube wall is absorbed in a thin gas layer near the wall.
Since the gas emission is isotropic, about one-half of the energy reemitted
by this thin layer goes back toward the wall. Thus the gas in the center
of the tube is shielded from the direct radiation and the heat transfer
efficiency decreases.

7.4.3 Other Multimode Problems

Radiation effects in rocket exhaust plumes have been examined by
deSoto (ref. 42). Kadanoff (ref. 4) treated radiation effects in ablating
bodies. A very large body of literature exists that deals with reentry of
bodies into the atmosphere (ref. 43) and radiation within and from
hypersonic shocks. A rigorous treatment of these problems is difficult
because of the nonequilibrium chemical reactions that are coupled with
the radiation effects. References 43 and 22 to 25 give a good introduction
and discussion of shock problems. Radiation interactions with a layer of
gas including transpiration is treated in reference 44.


The treatment of multimode problems in absorbing-emitting media
can be viewed as solving conduction or convection problems with a
distributed energy source (or sink) term present. This source (or sink)
term is the local net radiation gain (or loss) within the medium. The
radiative source tern1 can be determined by either exact or whatever
approximate methods can be applied in a given situation. Cases were
mentioned in this chapter where exact, diffusion, additive, and optically
thin formulations were used. As with almost all radiation problems,
the basic equations describing the physical processes can usually be
written. The difficulty is in solving these equations. It is not possible to
discuss in detail the wide variety of physical situations and solution

methods that can b e f o u n d in t h e l i t e r a t u r e . A number of references

have been provided in t h i s c h a p t e r to help guide t h e r e a d e r to s o m e of
t h e p e r t i n e n t basic i n f o r m a t i o n .

1. KELLETT,B. S.: Transmission of Radiation Through Glass in Tank Furnaces. J . Soc.
Glass Tech., vol. 36, 1952, pp. 115-123.
2. GARDON, ROBERT:The Emissivity of Transparent Materials. J. Am. Cer. Soc., vol. 39,
no. 8, Aug. 1956, pp. 278-287.
3. CONDON, EDWARD U.: Radiative Transport in Hot Glass. J. Quant. Spectrosc. Radiat.
Transfer, vol. 8, no. 1, 1968, pp. 369-385.
4. KADANOFF, LEO P.: Radiative Transport Within an Ablating Body. J. Heat Transfer,
vol. 83, no. 2, May 1961, pp. 215-225.
5. MERRIAM,R. L.; A N D VISKANTA,R.: Radiative Characteristics of Cryodeposits for
Room Temperature Black Body Radiation. Presented at the Cryogenic Engineering
Conference, Case-Western Reserve Univ., Cleveland, Ohio, Aug. 1968.
6. MCCONNELL, DUDLEYG.:Radiant Energy Transport Within Cryogenic Condensates.
Presented at the Institute of Environmental Sciences Annual Meeting and Ecpipment
Exposition, San Diego, Calif., Apr. 11-13, 1966.
7. VISKANTA,R.; AND GROSH,R. J.: Heat Transfer by Simultaneous Conduction and
Radiation in an Absorbing Medium. J. Heat Transfer, vol. 84, no. 1, Feb. 1962, pp.
8. VISKANTA, R.; A N D GROSH,R. J.: Effect of Surface Emissivity on Heat Transfer by
Simultaneous Conduction and Radiation. Int. J. Heat Mass Transfer, vol. 5, Aug.
1962, pp. 729-734.
9. VISKANTA,R.; AND MERRIAM,R. L.: Heat Transfer by Combined Conduction and
Radiation Between Concentric Spheres Separated by Radiating Medium. J. Heat
Transfer, vol. 90, no. 2, May 1968, pp. 248-256.
10. GREIF,RALPH;AND CLAPPER,GEANP.: Radiant Heat Transfer Between Concentric
Cylinders. Appl Sci. Res., Sec. A, vol. 15, 1966, pp. 469-474.
11. TIMMONS,D. H.; AND MINGLE,J. 0.: Sir~tultaneousRadiation and Conduction ~vith
Specular Reflection. Paper 68-28, AIAA, Jan. 1968.
12. EINSTEIN,THOMASH.: Radiant Heat Transfer to Absorbing Gases Enclosed Between
Parallel Flat Plates with Flow and Conduction. NASA TR R-154, 1963.
13. CESS, R. D.: 'The Interaction of Thermal Radiation With Conduction and Convection
Heat Transfer. Advances in Heat Transfer. Vol. I. Tllomas F. Irvine, Jr., and James
P. Hartnett, eds., Academic Press, 1964, pp. 1-50.
14. HOWELL,JOHNR.: Detern~inationof Combined Conduction and Radiation of Heat
Through Absorbing hledia by the Exchange Factor Approximation. Chem. Eng.
Progr. Symp. Ser., vol. 61, no. 59, 1965, pp. 162-171.
15. GOLDSTEIN, MARVIN;AND HOWELL,JOHNR.: Boundary Conditions for the Diffusion
Solution of Coupled Conduction-Radiation Problems. NASA TN D-4618, 1968.
16. HOWELL,J. R.; A N D GOLDSTEIN,M. E.: Effective Slip Coefficients for Coupled Con-
duction-Radiation I'roblems. J. Heat Transfer, vol. 91, no. 1, Feb. 1969, pp. 165-166.
17. TAITEL,YEHUDA; AND HARTNETT,J. P.: Application of Rosseland Approximation and
Solution Based on Series Expansion of the Emission Power to Radiation Proble~ns.
AIAA J., vol. 6, no. 1, Jan. 1968, pp. 80-89.
18. WANG,I,. S.; AND TIEN,C. L.: A Study of Various Li111its in Radiation Heat-Transfer
Problerns. Int. J. Heat Mass Transfer, vol. 10, no. 10, Oct. 1967, pp. 1327-1338.

19. NESS, A. J.: Solution of Equations of a Thermal Network on a Digital Computer. Solar
Energy, vol. 3, no. 2, 1959, p. 37.
20. LICK, WILBERT:Energy Transfer by Radiation and Conduction. Proceedings of the
1963 Heat Transfer and Fluid Mechanics Institute. Anatol Roshko, Bradford Sturte-
vant, and D. R. Bartz, eds., Stanford University Press, 1963, pp. 14-26.
21. VISKANTA,R.: Radiation Transfer and Interaction of Convection with Radiatioll
Heat Transfer. Adva~lcesin Heat Transfer. Vol. 3. Thomas F. Irvine, Jr., and Ja~lles
P. Hartnett, eds., Academic Press, 1966, pp. 175-251.
22. PAI, SHIH-I: Radiation Gas Dynamics. Springer-Verlag. 1966.
Theory of Gas Dynamics. Addison-Wesley Publ. Co., 1965, chs. 10-13.
24. ZEL'DOVICH, YA. B.; AND RAIZER,Yu. P.: Physics of Shock Waves and High-Tempera-
ture Hydrodynamic Phenomena. Vol. I , Academic Press, 1966, Pt. 11.
25. VINCENTI,WALTERG.; AND KRUGER,CHARLESH., JR.: Introduction to Physical Gas
Dynamics. John Wiley & Sons, Inc., 1965, chs. 11-12.
26. NOVOTNY,J. L.; AND YANG,KWANG-TZU:The Interaction of Thermal Radiation in
Optically Thick Boundary Layers. J. Heat Transfer, vol. 89, no. 4, Nov. 1967, pp.
27. VISKANTA, R.; AND GROSH,R. J.: Boundary Layer in Thermal Radiation Absorbing and
Emitting Media. Int. J. Heat Mass Transfer, vol. 5, Sept. 1962, pp. 795-806.
28. CESS, R. D.: Radiation Effects Upon Boundary-Layer Flow of an Absorbing Gas. J.
Heat Transfer, vol. 86, no. 4, Nov. 1964, pp. 469-475.
29. FRITSCR,C. A,; GROSH,R. J.; AND WILDIN,M. W.: Radiative Heat Transfer Through
an Absorbing Boundary Layer. J. Heat Transfer, vol. 88, no. 3, Aug. 1966, pp. 296-
30. HOWE, JOHN T.: Radiation Shielding of the Stagnation Region by Transpiration of
an Opaque Gas. NASA TN D-329, 1960.
31. NOVOTNY,J. L.; AND KELLEHER,M. D.: Free-Convection Stagnation Flow of an
Absorbing-Emitting Gas. Int. J. Heat Mass Transfer, vol. 10, no. 9, SeW. 1967, pp.
32. CESS, R. D.: The Interaction of Thermal Radiation with Free Collvection Heat Transfer.
Int. J. Heat Mass Transfer, vol. 9, no. 11, Nov. 1966, pp. 1269-1277.
33. GILLE, JOHN;AND GOODY,RICHARD:Convection in a Radiating Gas. J. Fluid hIech.,
vol. 20, pt. 1, Sept. 1964, pp. 47-79.
34. CESS, ROBERTD.: The Interaction of Thermal Radiation in Boundary Layer Heat
Transfer. Third International Heat Transfer Conference. Vol. 5. AIChE, 1966, pp.
35. SCHLICHTING, HERMANN (J. KESTIN,TRANS.):Boundary Layer Theory. Fourth ed.,
McGraw-Hill Book Co., Inc., 1960, 11. 116.
36. VISKANTA, R.: Interaction of Heat Transfer by Conduction, Convection. and Radiation
in a Radiating Fluid. J. Heat Transfer, vol. 85, no. 4, Nov. 1963, pp. 318-328.
37. EINSTEIN,THOMASH.: Radiant Heat Transfer to Absorbing G a s e s Enclosed in a
Circular Pipe with Conduction, Gas Flow, and Internal Heat Generation. NASA
TR R-156, 1963.
38. ADRIANOV, V. N.; AND SHORIN,S. N.: Radiative Transfer in the Flow of a Radiating
Medium. Trans. TT-1, Purdue University, Feb. 1961.
39. CHEN,JOHNC.: Simultaneous Radiative and Convective Heat T r a n s f e r in an Absorb-
ing, Emitting, and Scattering Medium in Slug Flow Between Parallel Plates. Re!].
BNL-6876-R, Brookhaven National Lab., Mar. 18, 1963.

40. DESOTO,SIMON;AND EDWARDS, D. K.: Radiative Emission a r ~ dAbsorption in Noniso-

thermal Nongray Gases in Tubes. Proceedings of the Proc. 1965 Heat Transfer and
Fluid Mechanics Institute. A. F. Charwat, ed., Stanford Univ. Press, 1965, pp. 358-
41. LANDRAM, C. S.; GREIF, K.; AND HABIB.I. S.: Heat Transfer in Turbr~lentPipe Flow
with Optically Thin Radiation. J. Heat Transfer, vol. 91, no. 3, Aug. 1969, pp. 330-336.
42. DESOTO,SIMON:Tlie Radiation from an Axisymmetric, Real Gas System with a Non-
isothermal Temperature Distribution. Chem. Eng. Progr. Symp. Ser., vol. 61, no. 59,
1965, pp. 138-154.
43. PENNER,S. S.; AND OLFE, D. B.: Radiation and Reentry. Academic Press, 1968.
44. VISKANTA,R.; AND MERRIAM,R. L.: Shielding of Surfaces in Couette Flow Against
Radiation by Transpiration of an Absorbing-Emitting Gas. Int. J. Heat Mass Transfer,
vol. 10, no. 5, May 1967, pp. 641-653.
Chapter 8. Radiative Transfer in Scattering and
Absorbing Media
The extinction coefficient K for thermal radiation is composed of an
absorption coefficient and a scattering coefficient as discussed in section
1.5.1. The equation of transfer when only absorption is present was
treated in chapter 2. In this chapter extension is made to treating radia-
tive transfer for cases where the scattering phenomenon becomes im-
portant. The elastic scattering processes are considered that are impor-
tant insofar as radiative transfer calculations are concerned. Effects of
polarization and the various inelastic scattering processes are only
mentioned in passing. These processes along with other scattering effects
are found in the more comprehensive treatments of scattering problems
in references 1 to 6. The work of Van der Hulst (ref. 1) is especially
valuable in giving a detailed discussion of elastic scattering from single
When radiation impinges upon any substance, a part of the energy
i s removed by absorption, and another part is redirected by scattering.
Scattering may occur from particles or objects of any size from elec-
trons to planets, and scattering from particles or objects in each size
range can be important in special situations. For elastic scattering the
photon energy and therefore the frequency of the photon is unchanged
by the scattering; for inelastic scattering the photon energy is changed.
In theoretical developments scattering is usually considered for a
single particle. When a cloud of many particles is dealt with, the scat-
tering intensities from the individual particles are usually added, thereby
assuming that each particle scatters independently. This i s a valid
assumption if the particles are more than a few diameters apart. I n
most practical situations this assumption can be made as the particles
are separated by much larger distances.
There are various phenomena that may occur when incident radiation
strikes a particle. Some of the incident radiation may be reflected
from the particle surface. The remaining portion of the radiation will
penetrate into the particle where part of the radiation can be absorbed.
If the particle is not a strong internal absorber, some of this radiation will
pass back out. This may occur after traveling only a single path through
the particle or the radiation may undergo multiple internal reflections and
travel about within the particle before escaping. When interacting with
the particle boundary, the radiation will be refracted and will also have

its di~ectionchanged by subsequent intellla1 ~eflections.The rediiection

by these processes of the energy penetlating into the particle and then
escaping is termed scattering by refraction. Additional scattering is
caused hy diffraction which produces, for exanlple, the inteiference
patterns observed when light passes through an aperture in a screen.
Diffraction is the result of slight bending of the radiation propagation
paths when passing near the edges of an obstruction.
The reflection, refraction, and diffraction depend on the optical
properties (i.e., the complex refractive index f i = n - i ~ )of the par-
ticle, and of the size of the particle relative to the wavelength of the
incident radiation. An additional conlplication is the particle geometry.
It is usually assumed that the medium surrounding the particles has
a unity simple refractive index n and a zero extinction coefficient K so
that f i = 1 1 - i0= 1. In this instance the surrounding medium does not
enter into the optical behavior of the medium-particle system.
In principle the scattering behavior can be obtained from the solution
of the Maxwell electromagnetic equations that govern the radiation
field for the medium-particle system. However, the solution provides
very complicated relations for even simple particle geometries. Hence
in many instances a number of simplifications are made as will now
be outlined.
One simplification is the geometric one of letting the scattering
particles be spheres. This is not as restrictive an assumption as it might
appear since, as discussed in reference 1, the results for spheres do have
a wider geometric applicability. Consider an array of irregularly shaped
particles, the surfaces of which are assumed composed of convex portions
(no concave indentations). Because the particles are in a random orienta-
tion, an equal portion of surface elements will face each angular direction,
which is the same angular distribution of surface elements as for a spheri-
cal particle. The net result is that the angular distribution of scattered
radiation viewed at a distance from the actual particles will be the same
as that scattered from spherical particles.
A second simplification is to consider the limiting solutions for scatter-
ing from large and from small spheres. A convenient parameter is 71-Dlh
where D is the sphere diameter. For large spheres (mD/h greater than
about 5) the scattering is chiefly a reflection process and hence can be
calculated from relatively simple geometrical reflection relations. There
is also diffraction of the radiation passing near the sphere, but this is
accounted for separately as will be discussed in section 8.4.4. For small
spheres (mDIX less than about 0.6), the approximation of Rayleigh scatter-
ing can be used as will be discussed in section 8.4.5. For the intermediate
range of mD/X, the general Mie scattering results apply, but the results
of this general solution of Maxwell's equations are quite complicated.

A third type of simplification is to look at limiting cases of the optical

constants of the particle. For metals the n and K are often large so the
case can be considered where f i = n - i~ + m. For a dielectric ( K = 0 ) a
limiting case is where n = 1. In this instance the reflectivity of the
surface will be small.
Limiting types of surface conditions are also considered (i.e., specular
and diffuse surfaces). The particle surface can only act diffuse, however,
if its dimensions are large compared with the wavelength of the incident
The theory and results will be discussed for several of the more useful
scattering relations.

absorption coefficient
speed of light in medium
speed of light in vacuum
particle diameter
emissive power; electron charge
function of fi in Rayleigh scattering relation, eq. (8-23)
distance defined in connection with eq. (8-54)
radiative source function, eq. (8-51)
radiation intensity
Bessel function of first kind of order one
extinction coefficient, a a,
extinction mean free path
mass of electron
number density, particles per unit volume
simple refractive index
complex refractive index, n - i~
energy per unit time
energy flux, energy per unit area and time
classical electron radius
coordinate along path of radiation
scattering cross section
absolute temperature
parameter in eq. (8-21)
coordinate direction normal to plane layer
cone angle, angle from normal of area
circumferential angle

optical depth; extinction coefficient in complex refrac-

tive index
cos p
scattering coefficient
phase function for single scattering
scattering angle measured from forward direction to
direction to observer
spherical angle in fig. 8-4
albedo for single scattering, eq. (8-48)
solid angle
b black
D at position coordinate D
i incident
P particle; projected
S scattered or scattering
X spectrally dependent
+- propagating in direction having positive cos P
propagating in direction having negative cos P
1,2 refers to surfaces 1 or 2

directionally dependent quantity

dummy variable of integration
zeroth-, first-, or second-order term or moment


8.3.1 The Scattering Cross Section
The extent of scattering to be expected is often measured in terms of
the scattering cross section s . This is the apparent area that a n object pre-
sents to an incident beam insofar as the ability of the object to deflect
radiation from the beam is concerned. It is usually given in square
centimeters for thermal radiation properties. This apparent area may
be quite different from the physical cross-sectional area of the scatterers
as can be seen from some of the approximate cross sections in table
&I. In addition to depending on the particle size, the scattering cross
section may depend upon the shape and material of the scattering body,
I I I I -7
I Body I Physical cross Conditions Type of scattering Scattering cross
section. cm2 I section. cm2

I Photon I ..................... 1 Energy of incident photon small I ................................. / - 2x10-56 I

Free Energy of photon < electron kinetic energy Thomson
-- CT

Energy of photon + electron kinetic energy Compton

Atom or 0.88X 10-'6 Elastic. Energy of incident photon < Rayleigh Proportional to
molecule (first Bohr - 6.65X10-25/A4
electron molecule
orbit) or atom Energy of incident photon .>
Rayleigh, approaches
electronic binding energy Thomson

Inelastic Energy of incident photon 4 Raman - 6.65x10-25

electronic binding energy

Energy of incident photon S Approaches Compton * 3

electronic binding energy 1 8 :,,(;;+1n2el, 1 1
Particles of A %- D, single scattering / Rayleigh / Proprtional to Y'/A' 1
diameter D 4 I I
A - D Mie Varies widely 1
A<D Fraunhofer and
Fresnel diffraction
plus reflection

*r,= classical elertron radius. 2 , 8 1 8 ~10-13 cm: e=hv/m,c: where h is Planck's constant: nv= cross section for Tho~nsonscattering: In,.= r,lec~ronmass. 9.108X 10-'9.

and the wavelength, polarization, and coherence of the incident radiation.

The scattering cross section can be determined experimentally
by measuring the amount of radiation in a beam that is able to pene-
trate through a cloud of scattering particles. One experimental diffi-
culty is in separating the radiation that is scattered into the forward
direction from the radiation that is transmitted without any particle
interaction. This difficulty can be diminished by using an incident
beam with a very small divergence angle. Then the forward direction
of the transmitted radiation will encompass only a small solid angle
which will include only a small portion of scattered radiation. The ratio
of the scattered portion di;, ,of the incident intensity to the intensity i;
of the incident beam is equal to the ratio of the apparent pojected scat-
tering area d A s ~occupied by all scattering particles, to the cross-sec-
tional area of the incident beam dA. This gives the following for a beam
traveling a differential distance within a medium in which it encounters
the scattering area dA,i:

Note that the apparent projected scattering area of the particles can
and usually will depend on wavelength.

The apparent scattering area presented by a group of t h e scattering

particles is related to the average scattering areas of t h e individual
particles by

where Ns is the number density of the particles, sA is the average scatter-

ing cross section of the particles, and dV is a differential volume of the
particle containing cloud as shown in figure 8-1. Inserting equation
(8-2) into equation (8-1) gives the change dii of the intensity as a result
of scattering from the incident beam

There is also intensity scattered into the S direction which will contribute
to di; but this will be incorporated later.

By integrating equation (8-3) over a path from 0 to S , t h e intensity is

found at S as a result of attenuation by scattering from t h e beam with
original intensity i;(O),

Scattered intensity for

(cp, 0) direction
dii, s(p, 0)

-- Forward

attered radiation

8-1.-Scattering of intensity into direction (p, 0) from incident radiation within
solid angle dmi.

i; (S) = i; (0) exp (- 1S A N ~*)~ S (8-4 a)

The portion of the incident intensity that was scattered away along the
path is thus

i; (0) - i; (s) = i; [
(0) 1 - exp (- % SAN,~S*)] (8-4b)

The scattering coefficient r s h is now defined to be

so that equation (8-4a) becomes

i;(s) = i;(O) exp [- [f l S A ( s * ) d ~ * ] (8-6)

This is the pure scattering form of Bouguer's law (see section 1.5).
If there is a distribution of particle sizes to be considered in detail, the
preceding analysis can be generalized. Let Ns(R)dR be the 'number of
particles per unit volume in the radius range from R to R + dR and let
sx(R) be the scattering cross section for a particle of radius R. Then by

integrating over all the particles, the scattering coefficient is

As in the interpretation of the extinction coefficient (section 1.5.2),

the scattering coefficient msk can be regarded as the reciprocal of the
mean free path that the radiation will traverse before being scattered.
The a s k is thus a reciprocal length, and can be regarded as a scattering
area per volume along the path, a , ~ = d A , k / d V from equations (8-5)
and (8-2). Since sh is mSk/Ns,it is the effective scattering area per particle.
The ratio of sh to the actual geometric projected area of the particle
normal to the incident beam is termed the eficiency factor.
At particle densities near or below the molecular density of air at
1 atm (Ns= 2.7 X 1019 particles/cm3), it can be seen that for most of
the processes listed in table 8-1 the scattering coefficient will be very
small (and thus the scattering mean free path very long). This is especially
true for photon-photon, Thomson, and Raman scattering, which may
generally be ignored in engineering radiative transfer calculations.
The previous relations have been concerned with what portion of
the intensity in an incident beam is lost as a result of being scattered
away along a path. The additional information that will be required to
formulate radiative transfer relations for scattering media is the direc-
tional distribution of the scattered radiation. This is given in terms of an
angularly dependent phase function.

8.3.2 The Phase Function

Consider the radiation within solid angle dwi that is incident on area
dA in figure 8-1. The entire portion of the incident intensity that is
scattered away in the distance dS is given by equations (8-3) and (8-5)
dii, ,= mshiidS (8-8)

The dii, ,is the spectral energy scattered within path dS per unit incident
solid angle and area normal to the incident beam

As shown by figure 8-1 the scattered energy produces an intensity dis-

tribution as a function of angles 8 and cp measured relative t o the forward

direction. A phase function cD(cp, '8) will be defined to describe the

scattered angular distribution.
The scattered intensity in any direction ( c p , 8 ) is defined as the
energy scattered in that direction per unit solid angle of the scattered
direction and per unit area and solid angle of the incident radiation;
that is,

spectral energy scattered in direction ( c p , 8 )

diL, ,(cp, 8 ) =

The directional magnitude of di;,,(cp, 8 ) is related to the entire

intensity di;, , scattered away from the incident radiation by the phase
function, such that

To better understand the phase function, note that the spectral energy
per dA scattered into dws per unit dwi and unit dA is d i i , ,(cp, 8)dws. Then
the spectral energy per unit dwi7 unit d A and dA scattered into all dw, is

di:, ,( 9 , 8)dws. However. the scattered energy per unit d o i , d A ,


and dX is di;, ,so that

Using equation (8-11) results in

which gives the phase function as

Thus, @(cp, 8 ) has the physical interpretation of being the scattered

intensity in a direction divided by the intensity that would be scat-
tered in that direction if the scattering were isotropic. For isotropic

scattering then, @ = 1. By integrating equation (8-13) over all dws, it

is evident that a(?,0 ) is a normalized function such that

The phase function can be a complicated function of cp and I3 as will

be shown in subsequent sections.


8.4.1 A Cloud of Large Specularly Reflecting Spheres
One of the most simple scattering configurations is a cloud of large
spherical particles (n-D/A > - 5) that have specularly reflecting sur-
faces. Figure 8-1 shows a differential volume element of the cloud with
cross section d A normal to the incident radiation and with thickness '

dS. The incident energy intercepted by the volume element is iidwidA.

It is assumed that the particle density is low enough so that each particle
scatters independently and there is negligible shadowing of the particles
by each other. Let the projected area of a particle normal to the direction
of ij, be A , so that the fraction of the incident energy on d A that strikes
the particle is A ,/dA. Part of this energy will be absorbed and the re-
mainder will be scattered by being reflected specularly.
The details of the reflection process are shown in figure 8-2. The
energy intercepted by a band of cross section RdP on the surface of the
sphere is equal to the energy intercepted by the particle multiplied by
Abrrrj(t/A,,,where Abnr,,jis the band area projected normal to i;. This gives

A,, Abnrrd
Energy intercepted by band= iidwidAdA - ---
dA A,,
= iidwidh 277.R2,sin P cos /3 d p

The amount of reflected energy is i;dwidA 2n-RGin P cos P d p p ; ( P )

where p; ( p ) is the directional specular reflectivity for incidence at angle
p. The amount of energy reflected from the entire sphere is found by
integrating over the sphere area, that is,

Reflected energy =i,'dwidA n-R2

r 2p;(P) sin P d ( s i n P )

From equations (3-43a) and (3-17a) of volume I the integral is the

hemispherical reflectivity ph. Hence the energy scattered by reflection


-Band on surfa

8-2.-Reflection of incident radiation by surface of specular sphere.

from the entire sphere is i;dwidA vR2 p i . B y using the scattering cross
section S A for the particle, the scattered energy is i;dwidA S A . Hence

and the particle scattering cross section is

Thus s~ is equal to the projected area of the particle times the hemispheri-
cal reflectivity. Inserting equation (8-15) into equation (8-7) gives the
scattering coefficient as

If all the spheres are the same size with radius R , it is not necessary to
integrate over the particle size distribution and the scattering coefficient
ash= p l r R 2 N s (8-17)

which is of the same form as equation (8-5).

To obtain the phase function, figure 8-2 shows that the energy spec-

ularly reflected from the band of the sphere at angle j3 will be reflected
into direction 2P and into a solid angle

dw, = 2?r sin 2j3 d ( 2 P )= 877 sin P cos j3 dj3

The phase function is concerned only with the portion of the energy that
is scattered. Since each particle is assumed to scatter independently, the
scattered portion of the radiation emerging from a portion d V of the par-
ticle cloud when observed at a distance large compared with the indi-
vidual particle diameter will have the same phase function as for a single
particle. The energy incident on a single particle is i;dwidXA,. Using the
scattering cross section in equation (8-IS), all the energv scattered away
by a particle is zidwidk ?rkLp ~ Then
. the incident intensity that is scat-
tered is from equation (8-9)

The energy scattered away by a particle into dw, is

iLdwidA 2 r R 2 sin j3 cos j3 d p p ; ( P )

The scattered intensity into direction 2j3 (where this intensity is defined
in eq. (8-10)) is

iidwidA 2?rR2 sin P cos j3 dj3 p i ( P ) ii;pi;(P)

d i i , ,(2j3) = -
dwiApdwsdA 4,n

Inserting this into equation (8-11) gives

The angle 2 6 is related to the angle cp in figure 8-2 by cp= r - 2 P so that

relative to the forward scattering direction

For unpolarized incident radiation the reflectivity p ; ( P ) for a dielec-

tric sphere can be found from equation (4-61) of volume I. Also, the direc-
tional-hemispherical reflectivity is equal to unity minus the emissivity
values in figure 4-5 of volume I. As shown by this figure, the p;(j3) for

8-3.-Scattering diagram for specular reflecting sphere that is large co~npared
with wavelength of incident radiation.

normal incidence is usually quite small compared with that at grazing

angles (d-1 at P = 90'). Consequently, the forward scatter from the
sphere (at cp=O) is unity, and the backward-scatter (at cp=7~)is small. By
use of figure 4-5 of volume I, the quantity PA@(p) can be plotted for vari-
ous indices of refraction n as shown in figure 8-3. The PAfor a dielectric
can be found by use of figure 4-6 of volume I.

8.4.2 Reflection from a Diffuse Sphere

For the specularly reflecting sphere in figure 8-2 the energy scattered
in each direction resulted from the reflection of energy at a single
location on the sphere. If the sphere is diffuse, however, each surface
element that intercepts incident radiation will reflect energy into the
entire 2 7 ~solid angle above that element. Thus the radiation scattered
into a specified direction will arise from the entire region of the sphere
that receives radiation and is also visible from the specified direction. This
is illustrated by figure 8-4(a). The shaded portion of the sphere will not
contribute radiation in the direction of the observer because it either
does not receive radiation or is hidden from the direction of observation.
Consider the sphere of radius R in figure 8-4(b). A typical surface
area element dA is located at angles and 6. The observer i s at angle cp
measured from the forward direction. The normal to dA is at angles

,Not illuminated

To observer

(b) observer

(a) Illuminated region visible to observer.

(b) Geometry on sphere.

8-4.-Scattering by reflection fro111diffuse sphere.

p and a relative to the directions of incidence and observation. The inci-

dent spectral energy flux within the incident solid angle dwi is iidwidh.
The projected area of dA normal to the incident direction is dA cos P, so the
energy received by dA is iLdwidhdA cos P. The amount of this energy
that is reflected is pLiLdwidXdA cos P , where p: is the diffuse directional-
hemispherical spectral reflectivity. The p: is assumed independent of
incidence angle and hence is equal to the hemispherical reflectivity pi.
Using the cosine law dependence for diffuse reflection gives the re-
flected energy per unit solid angle dw, in the direction of the observer
as pAi;dwidXdA cos /3 cos a / r . In order to integrate the reflected contri-
butions that are recieved by the observer from all elements on the sphere
surface, the dA, cos P, and cos a are expressed in terms of the spherical

coordinates R , $, and 6 which gives dA = R2 sin 6 d0d$, cos P= sin 0

cos $, and cos a = sin 6 cos ($ n- - p ) . Then the energy scattered by
reflection into the cp direction per unit solid angle dw, about that direc-
tion is

pAi;doid), R
sin3 0 cos $ cos ($+ n- - cp)d$d0

By integrating this becomes

p~iidwidAR 2 2
- (sin cp - cp cos cp)
IT 3

The energy per unit dA scattered in direction cp per unit dws and per
unit area and solid angle of the incident radiation is obtained by dividing
the scattered energy by nR2dwidA giving

di'A , ( S )~ = 2 -
T 2P
3 (sin cp - cp cos cp )

8-5. -Scattering phase function for diffuse reflecting sphere, large compared with
wavelength of incident radiation and with constant reflectivity.

The entire amount of the incident intensity that is scattered is di;,s=pt,iL.

Then from equation (8-ll), the directional magnitude of the scattered
intensity is equal to the entire scattered intensity times the phase

so that the phase function for a diffuse sphere is

@ ( q )=- (sin cp-q cos c p ) (8-20)

The @ ( q ) from equation (8-20) is plotted in figure &5. The largest

scattering is for cp= 180°, that is, toward an observer back in the same
direction as the origin of the incident radiation. In this instance, the entire
illuminated surface of the sphere is observed.

8.4.3 Large Dielectric Sphere with Refractive index Close to Unity

For a large dielectric ( K = 0 ) sphere with refractive index n 1 the
reflectivity of the particle surface approaches zero. The incident radia-
tion can thus pass with unchanged amplitude into the sphere and there is
no scattering by reflection as in sections 8.4.1 and 8.4.2. With the ex-
tinction coefficient zero, the radiation will pass back out of the sphere
with unchanged amplitude. However, the velocity c=c,/n inside the
sphere medium is slightly less than that outside, so that radiation
passing through different portions of the sphere and hence through
different thicknesses will have different phase lags. The resulting inter-
ference of the waves passing out of the sphere yields a scattering cross
C D 2 [2-j$
s h =-
( 1 -cos W )
where W= 2 (.rrD/h)( n- 1 ) . Additional information for this situation is
given in reference 1.

8.4.4 Diffraction from a Large Sphere

For large spheres there is diffraction of the radiation passing in the
vicinity of the particle. The effects of the diffraction and reflection must
be added to obtain the total scattering behavior. Fortunately the dif-
fraction is predominantly in the forward scattering direction. This
means that the diffraction can be included in the radiative transfer

I Screen

1 Diffracted

@ sin (D

(a) Diffraction of radiation by hole.

(b) Phase function for diffraction from large sphere.
8-6.-Diffraction by hole or large spherical particle.

as if it were part of the radiation transmitted past the particle without

interacting with the particle. As a consequence, diffraction can often be
neglected when considering the energy exchange within a scattering
The most familiar form of &!Traction is when light passes through
a small hole or slit. As shown in figure 8-6(a) the result is a diffraction
pattern of alternate illuminated and dark rings or strips. If a spherical
particle is in the path of incident radiation, Babinet's principle states
that the diffracted intensities are the same as for a hole. This is a con-
sequence of fact that the hole and particle produce complementary
disturbances in the amplitude of an incident electromagnetic wave.
The energy diffracted by a spherical particle is thus the same a s that
diffracted by a hole of the same diameter. As a consequence, the entire
projected area of the sphere takes part in the diffraction process, and
the scattering cross section for diffraction is equal to the projected
area ~ 0 ' 1 4 . Since diffraction and reflection occur simultaneously, the
total scattering cross section can approach 2(.rrD2/4)when the sphere is
highly reflecting.
The phase function for diffraction by a large sphere is given in terms
of a Bessel function of the first kind of order one (ref. 1 , pp. 107 and

-This function is plotted in figure 8-6(b). Since the abscissa is (.rrD/A)

sin cp, for particles with large ~TD/Athe diffracted radiation lies within a
narrow angular region in the forward scattering direction. For small parti-
cles where rD/A is of order unity, the theory leading to equation (8-22) is
invalid and the general Mie scattering theory must be applied. The
integration to show that equation (8-22) satisfies equation (8-14) is
discussed in reference 7 (p. 398); it is only necessary to integrate over
small cp which simplifies the integration considerably.

8.4.5 Rayleigh Scattering

For many common situations, the scattering particles are consid-
erably smaller in diameter than the wavelength of the incident radiation
(D -G A). Scattering from such particles is termed Rayleigh scattering
after Lord Rayleigh, who examined this situation. Rayleigh scattering
is important in the atmosphere where the gas molecules are the scattering
particles. The cross section for Rayleigh scattering can be derived
from quantum theory or electromagnetic theory. Originally, Rayleigh

derived the functional dependence by dimensional analysis and arrived

at the following result:

where V is the volume of a particle and (;(TI) is an unknown function of

the complex refractive index of the scattering material. The important
result is that for Rayleigh scattering, the scattered ene+gy i n a n y direction,
is proportional to the inverse fourth power of the wavelength of the inci-
dent radiation. 'The inverse dependence upon the fourth power of wave-
length shows that the shorter wavelengths will be Rayleigh scattered
with a strong preference when the incident radiation covers a wavelength
Rayleigh scattering by the molecules of the atmosphere accounts for
the background of the sky being blue, and for the Sun becoming red in
appearance at sunset. The blue portion of the incident sunlight is at the
short wavelength end of the visible spectrum. Hence it undergoes strong
Rayleigh scattering into all directions, giving the sky its overall blue back-
ground. Without molecular scattering, the sky would appear black except
for the direct view of the Sun. As the Sun is setting, the path length for
direct radiation through the atmosphere becomes much longer than dur-
ing the middle of the day. In traversing this longer path more of the short
wavelength portion of the spectrum is scattered away from the direct
path of the Sun's rays. As a result, at sunset the Sun takes on a red color
as the longer wavelength red rays are able to penetrate the atmosphere
with less attenuation than the rest of the visible spectrum. If many dust
particles are present, a deep red sunset may be seen.
If particles with a very limited range of sizes are present in the atmos-
phere, unusual scattering effects may be observed. Following the erup-
tion of Krakatoa in 1883, the occurrence of blue and green Suns and
Moons was noted over a period of many years. This effect was attributed
to particles in the atmosphere of such a size range as to scatter only the
red portion of the visible spectrum. On September 26 of 1950, a blue Sun
and Moon were observed in Europe, a phenomenon believed due to finely
dispersed smoke particles of uniform size carried from forest fires burn-
ing in Canada. Scattering cross sections for Rayleigh scattering.-Equation
(8-23) gives only the functional dependence of the scattered radiation
on wavelength and particle volume, so that additional information is
needed for the particle scattering cross section and the angular dis-
tribution of the scattered intensity. Consider first small nonabsorbing
( K = 0 ) particles so that fi = n and D < -
0.2hln where A is the wave-

length in the particle material. The Rayleigh scattering cross section

for unpolarized incident radiation is found from more advanced theory
to be

Often Rayleigh scattering cross sections are given in terms of the polar-
izability ol, of the particles. This is a proportionality factor relating
the forces induced in the molecules to the external electromagnetic
field. Specifically it relates the dipole moment per unit volume (defined
as the polarization) produced in the material to the external field. For
the case under discussion here the polarizability is

so that equation (8-24) can be written

Scattering particles Restrictions Polarizability,


Electrons (Thomson Energy of incident photon is small,

scattering) hv 6 meci

Small dielectric
Particle diameter is small compared
with wavelength in medium
and in particle.
Spacing be:~veen particles is small com- -
Medium containing
small particles
pared with wavelength (<A). Particle di-
ameter is very small (D G A) compared
I 3
ti2 - 1
n2+ 2

with A in both medium and particle.

Spacing between particles > D.

Medium containing *
Spacing between particles is large ( A). In2-11
small particles Particle diameter is very small (D < A). 47rN

Medium containing Spacing between particles is large ( B A). JA-11

small particles Particle diameter is very small ( D 6 A). 2nN
The ii is close to 1.

* eA
-= classical electron radius, 2.818 X 10-'3cm.

In this more general form, cross sections for various particles that
follow the Rayleigh scattering relations can be introduced by substi-
tuting the requisite form for a,, into equation (8-26). Table 8-11 gives
some quantities for individual particles and particles in a nonpartici-
pating medium.
The actual scattering cross section for particles in a medium may
vary with h in a manner somewhat different from a l / X 4 dependence.
In air at standard temperature and pressure, for example, the restric-
tions are satisfied such that Rayleigh scattering from the gas molecules
should govern. However, the variation of refractive index with wave-
length causes the variation of the scattering cross section to depart
somewhat from the 1/X4 dependence. This is shown in figure 8-7 where

.2 .4 .6 .8 1.0
Wavelength, A, pm

FIGURE8-7. -Comparison of actual Rayleigh scattering cross section for air at standard
temperature and pressure with l / A 4 variation (from ref. 6).

'[ sfim
i p:l of refraction,

Extinction coefficient, K

8-8.-Rayleigh scattering cross section as a function of simple index of refraction
and extinction coefficient.

the actual scattering dependence on wavelength is compared with

When the particles are of a conducting material with a complex index
of refraction ii.= n - i ~ the
, scattering cross section has the following
form which is more general than that of equation (8-24)

Inserting i i = n - i~ and taking the square of the absolute value as indi-

cated give

For K = O this reduces to equation (8-24). The quantity S A / ( ~ ~ + V ' / A ~ )

from equation (8-28) is in figure 8-8 for various n and K values. Phase function for Rayleigh scattering. - For incident unpolar-
ized radiation, electromagnetic theory gives for Ra~leighscattering

---- Isotropic

8-9. -Pltase functions for Rayleigh and isotropic scattering.

This is independent of the circumferential angle 0.

A plot of the phase function for Rayleigh and for isotropic scattering
is given in figure 8-9. For Rayleigh scattering, the scattered energy is
directed preferentially along the forward direction of the incident radia-
tion, and also strongly back toward the source of the radiation.

8.4.6 Mie Scattering Theory

When the particles that cause scattering are not large as treated in
sections 8.4.1 to 8.4.3, and are not small enough to fall into the range that
is adequately described by Rayleigh scattering relations, recourse must
be taken to more complicated treatments. This is for the approximate
range (0.21~~)< (DIX) < 1 where A is the wavelength inside the particle
material. Gustav Mie (ref. 2) originally applied electromagnetic theory to
derive the properties of the electromagnetic field that arises when a
plane monochromatic wave is incident upon a spherical surface across
which the optical properties n and K change abruptly. As a consequence,
the energy absorption by the medium, the absorption by the scattering
particles, or both can be accounted for. The results of this theory apply
over the entire range of particle diameters. As might be expected, strong
polarization effects can be present. In certain cases, the phase function
becomes very complicated as illustrated by figure 8-10.
Van der Hulst (ref. 1) gives an excellent detailed treatment of the Mie
theory. The limiting cases of very small and very large particles are
examined, and working formulae for all ranges of size are presented. The

Phase function
-- - - -- ----- - Portion of phase function due to
perpendicularly polarized component

(c) (d)
(a) v D l A + 0; metallic sphere; (b) TD/A = 9.15; metallic sphere;
ii = 0.57 - 4.29 i. n = 0.57 - 4.29 i.
(c) v D / h = 10.3; lnetallic sphere; ( d ) &/A = 8; dielectric sphere;
ii = 0.57 - 4.29 i. n = 1.25.
FIGURE8-10.-Phase functions for Rtie scattering from metallic and dielectric spheres
(on arbitrary scales) from references 1 and 7.

cross sections and phase functions are discussed for dielectric and
metallic particles of various shapes including spheres a n d cylinders.
Some further work on absorbing particles, using detailed M i e scattering
theory, has been done by Plass (ref. 3).
One of the simpler results from the Mie theory is for small spheres.
The general Mie equations can be expanded into a power series in terms
of the parameter TDIA giving the scattering cross section as

The second term in the square bracket is the first correction to the
Rayleigh scattering relation which was valid for very small particles.
When nD/X is very small, this term drops out and the relation becomes
equation (8-27).
For small spheres the limit can also be considered where ii becomes
very large. The scattering particles will then be very highly reflecting
such as for a cloud of metallic particles. The result for this case cannot
be obtained by letting ii in equation (8-30) approach infinity. As ii
becomes large, the small part of the incident radiation that does pene-
trate the particle becomes almost totally internally reflected. This
creates standing waves within the particle which provide resonance
peaks in the scattering. The expansion used to obtain equation (8-30)
did not account for this behavior. In the limit for ii-+ the scattering
cross section for small spheres is

If, in addition to ii-+ m, the particles are so small that only the first
term in the bracket of equation (8-31) is significant, then for unpolar-
ized incident radiation the phase function is given by

8-11.-Phase function for scattering of unpolarized incident radiation from small
sphere with ii -+ a.

3 1
a(,) =,[(I -5 cos ,)?+ (cos p-;)2]

A polar diagram of this function is given in figure 8-11. This shows

that in contrast to Rayleigh scattering (fig. 8-9) the highly reflecting
particles produce a very strong scattering back toward the source.


Now that some of the fundamentals of scattering behavior have been
examined, the methods of using this information in radiative transfer
calculations can be treated. First, pure scattering problems are ex-
amined, and then the complete absorbing, emitting, and scattering case
is considered.

8.5.1 The Equation of Transfer in a Pure Scattering ~tmosphere

Let us consider first the situation of radiation transfer in a medium
where there is scattering but no absorption or emission of radiation.
The local intensity along a path will be attenuated by radiation scattered
out into other directions, and will be enhanced by radiation scattered into
the direction being considered. Figure 8-12 shows radiation with inten-
sity i; passing through a volume element dAdS where dA i s normal t~ the
direction of i;. While passing through the distance d S , a portion di;,,of
the intensity will be scattered away. From equations (8-3) and (8-5) this
is equal to
di;,,= i; (S) a,,idS (8-33)

-s \
8-12. -Scattering of energy into S direction.

TO compute the scattering from all directions into the direction of i:,
consider the radiation incident at angle ( p , 6 ) as shown in figure 8-12.
This radiation has intensity i ; ( P , O), and in the process of going through
the volume element dV will pass through a path length dS/cos P. From
equation (8-11) oriented with respect to the present coordinate system,
the intensity scattered from i ; ( P , 0 ) into the direction of ij( is

However, from equation (8-10) i;,, is an intensity defined as energy in the

scattered direction per unit dA, per unit scattered solid angle, per unit
incident solid angle h i , and per unit area normal to the incident intensity.
This is the area normal to i,'(P, 0 ) which is dA cos P . Then the spectral
energy scattered into the S direction as a result of i ; ( P , 0 ) is by use of
equation (8-34)

= c S A i(;p , e) A P
d ~ d ~ i d A dcos
cos p 477-

T h e contribution of this scattered energy to the spectral intensity in the

S direction is then

T o account for the scattering contributions from incident intensities

from all directions, integrate over all c l w i to obtain

d4Qi;, --
dAdwdA 4rr
i;(p,@)@(p,O)dwi (8-36)

T h e scattering particles have been assumed randomly oriented so that

the scattering cross section USA is independent of the incidence direction.
By combining equations (8-33) and (8-36) the change of intensity in
direction S is

AS in equations (2-5) and (2+) a scattering optical thickness KSA can be

introduced such that

~ K --~ csxdS

K ~ (AS ) =
I," ~ S (S*)
A dS *

Then equation (8-37) becomes

This has the same functional form as the equation for absorbing-emitting
media (eq. (2-7)). As in equation (2-7), the term on t h e right side of
equation (8-40) accounts for intensity added to the beam a t each point
along the path.
Equation ( 8 4 0 ) can be integrated along a scattering optical path from
0 to K S X in the same fashion as equation (2-10) was obtained. BY analogy
with equation (2-10) this gives

~;(K,A) = i i ( 0 ) exp (- K,A) + LKsA[k /w,-ir

i ; ( P , O)@(P1
x exp [- ( K ~-
A K$)] ~KZX (8-41)

where K.?A is a dummy integration variable, and the integral in the first
square bracket is evaluated at K&. For an ideal scattering p r o c e s s where
there is no absorption of photon energy by the scattering particles, there
will be no exchange of energy with the medium. In this i n s t a n c e , there is
only a directional redistribution of energy by the scattering. Then if
W S A and @ are independent of temperature, equation ( 8 4 1 ) applies

regardless of the temperature distribution of the medium.

In situations such as a searchlight or laser beam, the only significant

intensity source is that arising from the beam. Then the scattered
from other sources into the direction of the beam will be negligible and
equation (8-41) reduces to the simple exponential a t t e n u a t i o n relation

~ ; ( K ~ A=) i i ( 0 ) exp (- K ~ A ) (8-42)

8-1: Derive the equations that describe the l o c a l intensity

and energy flux in a one-dimensional scattering layer. The layer is con-

tained between infinite parallel black plates separated b) a distance D.
The lower and upper plates are maintained at temperatures T I and T2,
respectively. The gas scatters isotropically, is nonconducting, and has
a n absorption coefficient of zero. The gas properties are assumed
independent of wavelength.
Since the properties are assumed independent of wavelength, the
A subscripts can be omitted in what follows. The same relations will
apply, however, for radiation at a single wavelength if the spectral emis-
sion from the walls is utilized. Note that for isotropic scattering the phase
function is 1.
This example is the scattering counterpart of the development in
section 2.6.2 which was concerned with a nonscattering, absorbing-
emitting medium. If x is the distance measured normal from plate 1,
the scattering optical distance from a point on the plate to a point in the
medium is ~ ~ / c po where
s it should be noted that this K, is based on the
coordinate x, not on the actual path length of the radiation. Then by
analogy with equations (2-34) and (2-35) the intensities in the directions
having positive and negative cos p are respectively (see fig. 2-5)

- (K, - K:) IT
Xexp [ cos p ]dr: 0 ~ 82 ~ (8-43)

i l ( ~ ,p,) =- exp

coip [i

-- L*=.i'(~:, P*) sin p * d p * ]

K$ - K, IT
Xexp (------)d~: (8-44)
cos p
These relations for isotropic scattering can be solved in a manner
analogous to that for an absorbing-emitting medium. Note that the
integral in the last term on the right in equations (8-43) and (8-44) in-
volves integrating over all /3* directions and hence includes the contribu-
tions from both :i and il. A P * notation has been used to distinguish
this integration variable over all directions from the /? which has a
restricted range for each of the equations.
The energy flux flowing from wall 1 to wall 2 can be found by analogy
with equation (2-41) as

+ c~o s pj ~o ~ , ~ [ ~ / ~ @*I
* =sin ~p*dpX
i t ( ~ f , I
x exp (- cos p ) dK?} d~ (8-45)

In this equation the a T f is the energy flux leaving wall 1 by radiation. The
second term is the flux incoming to wall 1 in two ways: (1) along direct
paths from wall 2 along which the emitted radiation UTzJ is attenuated
by scattering, and (2)by radiation scattered in at local between
the plates and then attenuated before reaching wall 1.

8.5.2 Scattering in Absorbing-Emitting Media

If scattering, absorption, and emission are all of significance, then the
equation of transfer as given by equation (2-4) is - by including
the scattering terms of equation (8-37) to give
dS a ~ i ; (+aAi;,(s)
~) +=pi($
- v s A i ; ( ~ ) 47~. w;=4% oi)@(A, m, oi)dmi

Loss by ab- Gain by Loss by Gain by s c a t t e r i n g

sorption emission scatter- into S-direction
(including (not in- ing
the contri- cluding
bution by induced
induced emission)

The two terms representing losses by absorption and s c a t t e r i n g can be

combined. Then the equation of transfer for absorbing, emitting, and
scattering media (for the case of elastic anisotropic s c a t t e r i n g ) is the
following for the intensity in the solid angle o about the s - d i r e c t i o n :

The sum aA+cr,~ is the extinction coefficient Kk d i s c u s s e d in section

The albedo for single scatter 0, is sometimes used w h i c h is defined

as the ratio of the scattering coefficient to the extinction coefficient, or

The optical depth or opacity when both scattering and absorption are
present is given by (this has been defined previously in eq. (1-17))

where S* is a dummy variable of integration. Equation (8-47) now


Often, especially in the astrophysical literature, the final two terms in

equation (8-50) are combined into the source function 1; ( K * ) , defined as

This is the source of intensity along the optical path from both emission
and incoming scattering. The equation of transfer then becomes

The generalized equation of transfer including absorption, emission, and

scatter is thus quite similar in form to the equation of transfer for pure
absorption and emission that has been studied at length in chapters 2 and
3. Note that when f l o A - + 0 (no scattering), equation (8-52) does indeed
reduce to the correct form for pure emission and absorption (eq. (2-7)).
For f i o ~ - 1
+ (pure scattering), equation (8-52) reduces to the pure scat-
tering form (eq. (8-40)).
Because equation (8-52) is similar in form to equation (2-7), many
of the mathematical approaches to solutions of the equation of transfer
given in chapter 3 also apply when scattering is included. T h e texts
by Chandrasekhar and Kourganoff (refs. 4 and 5) and the work of Goody
(ref. 6) that deals with atmospheric effects treat at length scattering
problems with and without absorption and emission for one-dimensional

atmospheres. A demonstration of the similarity with previous derivations

given in chapter 3 will now be given by considering the diffusion

EXAMPLE8-2: Derive a first-order diffusion relation for radiative

transfer in a one-dimensional layer of absorbing-emitting isotropic
medium with isotropic scattering.
Let x be the thickness coordinate normal to the layer boundaries.
The equation of transfer for the general one-dimensional case including
emission, absorption, and scattering is from equation (8-52) (as in eq.

where p= cos p. For isotropic scattering the source function I: is inde-

pendent of direction and is given from equation (8-51) b y

In the diffusion approximation, the medium is optically dense. Con-

sequently the radiation arriving at any location comes only from the im-
mediate surroundings, as any other radiation would b e absorbed or
scattered before arriving at that location. Also in the diffusion approxi-
mation the radiant energy density changes slowly with position relative
to distances for attenuation. This can be stated more rigorously by letting
H be a path length over which the radiant energy density does
change appreciably, and letting I,, be the extinction m e a n free path,
1111= 11 (ah c S x ) . Then for the diffusion approximation to apply LllllH+ 1.
As in equation (3-19) the intensity is expanded in terms o f powers of the
small quantity

Insert equation (8-54) into the equation of transfer to o b t a i n


where I,,, ,= l / v s A . In addition to the expansion parameter I , , , / H , there

has appeared an additional quantity 1111/1,1,, , which gives the relation of
the total extinction to that of scatter alone. For the diffusion approxi-
mation including absorption and scattering, this parameter will have to
b e of order one-half. For small 1,,,/1,,,,,the problem would degenerate to
one of diffusion by absorption alone; this could be the situation in a dense
gas containing a few scattering particles. For l,l,/lnl,s approaching 1 so
= I l a , s , there would be scattering alone as in a thin carrier gas with
many scattering particles. Collect the terms in equation (8-55) of zeroth
order in I,,/H to obtain

The terms iLb and

Io i=4m
i;(O)dwi on the right in equation (8-56) do not

depend on the incidence angle dwi. Hence i;(O)on the left cannot depend
on angle. By using this fact, equation (8-56) reduces to

which further reduces to

i;(o)= Z.' A ~ (8- 57 )

Now collect terms from equation (8-55) of first order in II,,/H to obtain

Substitute equation (8-57) for iL(O)to give

T o find i?), multiplylby dwi = 271- sin P dp = - P r d p and integrate over

all solid angles

The integral on the left is zero so

Hence Lm i;(I)dwi= 0 and equation (8-58) reduces to

Substitute equations (8-57) and (8-59) into equation (8-54) to obtain

This is the same form as equation (3-23) for the nonscattering case,
except that the extinction coefficient aA+aSA appears i n place of the
absorptioil coefficient. Then by analogy with equation (3-25) the energy
flux in the x direction is given by the diffusion equation

The net flux depends only on the emissive power gradient and the ex-
tinction coefficient when diffusion conditions apply.

Bobco (ref. 8) used a modified diffusion solution to find the direc-

tional emissivity for radiation from a semi-infinite slab of isothermal
gray scattering-absorbing medium. The scattering was assumed iso-
tropic. The directional emissivities of the slab were found to differ con-
siderably from a diffuse distribution. Hsia and Love (ref. 9) treated
energy transfer between parallel plates through an anisotropically scat-
tering medium. Nonisothermal conditions in the medium were
considered. SoIutions were obtained by approximating t h e integral terms
in the equation of transfer by a finite summation. The s e t of differential

equations that resulted were solved by a specialized matrix transfor-

mation technique.
Love et al. (ref. 10) studied plane and cylindrical boundaries with
given reflectivities that enclose absorbing, emitting, and scattering gases.
Both Monte Carlo and discrete ordinate methods were used for com-
puting the energy transfer. Some experimentally determined values
of the scattering phase functions for glass beads, and aluminum, carbon,
iron, and silica particles were used for comparison of their effect in a
variety of energy exchange cases. It is significant that little difference
in energy transfer was found for the results using these experimental
phase fullctiolls as compared with the results using either the Rayleigh
or isotropic phase functions. It appears therefore that the assunzption
of isotropic scattering is often justified i n energy exchange calcuLations
i n enclosures.
In figure 8-13, results from reference 10 are shown for the fraction

Dimensionless radial position, rlR

8-13.-Effect of scattering phase function on energy scattered back to base plane
by cylinder of scattering medium. Optical diameter of cylinder, 2; height to diameter
ratio, 5 (ref. 10).

of emitted energy from a black disk that is scattered back to the base
plane from a cylinder of gas adjacent to the disk. The results for various
scattering phase functions in the gas are in very good agreement The
various phase functions gave energy transfer results that had less varia-
tion in plane ~ a r a l l e lgeometries than in the cylindrical geometry. It
should be emphasized that in some cases the insensitivity of results
to the scattering phase function is probably not a valid assumption.
Specifically the phase functioil will be important for beam transmission
or other situations where strong sources transmit directionally into a
scattering atmosphere.

1. HULST,HENDRIKC. VAN DER: Light Scattering by Small Particles. John miley & Sons,
Inc., 1957.
2. MIE,GUSTAV: Optics of Turbid Media. Ann. d. Physik, vol. 25, no. 3, Mar. 3, 1908,
pp. 377-445.
3. PLASS, GILBERT N.: Mie Scattering arid Absorption Cross Sections for Absorbillg
Palticles. Appl. Opt., vol. 5, no. 2, Feb. 1966, pp. 279-285.
4. CHANDHASEKHAR, SUBRAWMANYAN: Radiative Transfer. Dover Publ., 1960.
5. KOURCANOFF,VLADIMIR: Basic Methods in Transfer Problems; Radiative Equilibrium
and Neutron Diffusion. Dover Publ., 1963.
6. GOODY, R. 31.: Atmospheric Radiation. Vol. I. Theoretical Basis. Clarendon Press,
Oxford, 1964.
7. BORN,MAX; AND WOLF,EMIL: Principles of Optics. Seconded., Pergaolon P r e s s , 1964.
8. Bo~co,R. P.: Directional Emissivities from a Two-Dimensional, Absorbing-scatter-
ing Medium: The Semi-Infinite Slab. J. Heat Transfer, vol. 89, no. 4, Nov. 1967,
pp. 313-320.
9. HSIA, H. M.; AND LOVE,T. J.: Radiative Heat Transfer between Parallel P l a t e s Sepa-
rated by a Nonisothermal Medium with Anisotropic Scattering. J. Heat Transfer, vol.
89, no. 3, Aug. 1967, pp. 197-204.
10. LOVE, TOMJ.; ET AL.: Radiative Heat Transfer in Absorbing, Emitting and Scattering
Media. Oklahoma Univ. (ARL-67-0210, DDC No. AD-666427), Dec. 1967-
Chapter 9. Some Specialized Effects in Absorbing-
Radiating Media
This chapter deals with four special radiation topics each of which
has important applications, but in restricted areas. For this reason,
some of the features of these topics are only briefly presented.
The first topic deals with media having a nonunity refractive index.
This topic includes the radiation within, and penetrating into, materials
such as glass and ice. When radiation from one material enters one of
another refractive index, the bending and reflection of rays at the inter-
face must be considered in the analysis. For closely spaced layers
such as in cryogenic super insulation, there is an additional effect of
radiation tunneling between layers.
The second topic is flames, both nonluminous and those containing
luminous particles, mainly soot. A nonluminous hydrocarbon flame
contains carbon dioxide and water vapor as its chief radiating constitu-
ents. Radiation by these gases is fairly well understood. When soot is
present and the flame thereby becomes luminous, the radiation is
dependent on the radiative properties of the soot and the soot concentra-
tion within the flame. There is some information available on soot radia-
tive properties, but the amount is insufficient. In addition to the un-
certainties in the soot properties, a serious difficulty in flame radiation
computations is determining the soot concentration. The concentration
depends on the particular fuel, the flame geometry, and the complicated
mixing phenomena within the flame. At present there is no way of com-
puting soot concentration from the basic parameters, such a s the burner
geometry, fuel-air ratio, and the particular fuel.
This chapter ends with brief discussions of two topics: lun~inescence
and transient gas radiation problems. The recent. limited interest in
the latter stems chiefly from nuclear weapons calculations.

A area
a absorption coefficient
C volume fraction of particles in medium
C1, C2 constants in Planck's spectral energy distribution
C3 constant in Wien's displacement law

speed of light in medium

heat capacity at constant volume
ineail heat capacity
half thickness of slab; particle diameter
absorptio~lefficiency factor for single particle
emissive power
fuilctioll defined by eqs. (9-22) and (9-23)
radiation intensity
constants in equations for soot absorption
mean beam leilgth
number of moles
number of particles per unit volume
simple refractive index (real part of 7i= rz - i ~ )
partial pressure
energy per unit time
energy flux, energy per unit area and time
coordinate along path of radiation
absolute temperature
dimensionless time, (cruT;/pc,.)t
rectangular coordinate
exponent in eq. (9-18)
cone angle, angle from ilorrnal of area
angles giving total internal reflection
maximum angle where refraction occurs
dimensio~llesstemperature, TIT,
circumferential angle
optical thickness, ax; extinction coefficient in colllplex refrac-
tive index
Stefan-Boltzmann constant
scattering angle measured fro111forward direction to direction
of observer (see fig. 8-1)
solid angle

b blackbody
D evaluated for length D
e emitted
maximum value
no self-absorption; initial value
prod products
r red
ref reference value
vac in vacuum
A spectrally dependent
1,233 medium or boundary 1, 2, or 3
directionally dependent quantity


Most of the discussion in this volume has been concerned with dielec-
tric media that have a refractive index n of unity. This is not unduly
restrictive, as the absorbing-emitting medium is usually a gas, and
almost all gases have a refractive index that is very close to unity as
shown by table 9-1 (ref. 1).However, there are certain situations where
the refractive index can be significantly different from unity or can be
variable over a given path for radiation as a result of spatial temperature
variations. Table 9-1 lists some common materials that possess nonunity
refractive indices. As discussed in section 2.4.12 of volume I, one effect
of nonunity n is to increase the blackbody emission within the m,edium.
by a factor of n2. In this section, consideration is given to some situations
where the effect of refractive index must be considered.

9.3.1 Media With Constant But Nonunity Refractive Index

Consider the case of radiation with intensity i: in a dielectric medium
of refractive index nl. Let the radiation in solid angle dwl pass into a
dielectric medium of refractive index n2 as pictured in figure 9-1. As
a result of the differing indices of refraction, the rays will change direc-
tion as they pass into medium 2. The radiation in solid angle dwl at
incidence angle will pass into solid angle do2 at an angle of refraction
Pz If it is assumed that there is no reflection or scattering at the interface,
the energy of the radiation is conserved when crossing the interface.
From the definition of intensity this conservation of energy is given by

Material Refractive index, n 1

Air 1.00029
Argon 1.00028
Carbon dioxide 1.00045
Cljlorine 1.00077
Hydrogen 1.00014
Methane 1.00044
Nitrogen 1.00030
Oxygen 1.00027
Water vapor 1.00026


Chlorine 1.385
Ethyl alcohol 1.36 to 1.34 (16" to 76' C)
Oxygen 1.221
Water 1.33 to 1.32 (14" to 100' C)

I Solids I
Rock salt

Medium 1with refrac-

tive index nl

FIGURE9-1.-Beam with initial intensity i; crossing interface between two d i &lectric

media wit11 unequal refractive indices.

i: cos pl dAdwl = i; cos p2 dAd~2 (9-1)

where dA is an area element in the plane of the interface. Using the rela-
tion for solid angle

dw = sin p dPdO (9-2)

results in equation (9-1) becoming (noting that the increment of cir-

cumferential angle dB is not changed in crossing the interface)

i; sin j31 cos dP1= ii sin pz cos dpz (9-3)

From equation (4-43) of volume I, Snell's law relates the indices of

refraction to the angles of incidence and refraction by

- -
n1 - sin P2

nz sin Dl
Then by differentiation

Substituting equations (9-4) and (9-5) into equation (9-3) gives

Although equation (9-6) was derived for radiation crossing the interface
of two media, the equation also holds for intensity at any point in a trans-
parent lnediurn with variable refractive index so long as the local prop-
erties of the medium are independent of direction, that is, are isotropic.
This isotropy will be the case except in certain plasma physics applica-
tions. Thus, in general in a transparent isotropic medium, for either the
spectral intensity, or the total intensity in a medium with spectrally
independent refractive index

constant (9-7)

9.3.2 The Effect of Brewster's Angle

Consider a volume element dV inside a semi-infinite region of refrac-
tive index n z as shown in figure 9-2. Suppose that diffuse radiation of

FIGURE9-2.-Effect of refraction on radiation transport in media with nonunity refractive


intensity ii is incident upon the boundary of this region from a region

having refractive index n , ~
where nl < nn. Radiation incident at grazing
angles to the interface ( P I 90') will be refracted into medium 2 a t a
maximum value of j32 given by

sin P 2 , ,,,clt=- sin 90°=lkl
n2 n 2

Hence the volume element in medium 2 will receive direct radiation

from medium 1 only at angular directions within the range

Now consider e ~ i s s i o nfrom dV. The portion of this emission t h a t

enters region 1 will be along paths found by reversing the arrows o n the
solid lines in figure 9-2. However, there is also radiation from dV a l o n g
paths such as those shown by the dashed lines in figure 9-2 that are
incident on the interface at angles P,*, where

sin pz > -

From equation (9-4), this means that such a ray would enter medium 1
at an angle given by

sin pl =% sin /32> ---

( = 1)
n~ n1n2

But sin pl cannot be greater than unity for real values of pl. This result
is interpreted to mean that any ray incident upon the interface from
medium 2 at any angle greater than that given by

cannot enter medium 1, and must be totally reflected at the interface.

The angle. defined by equation (9-12) is called Brewster's angle.
From section 2.4.12 of volume I, the blackbody emission inside a
dielectric medium with refractive index that is constant but not unity,
has an intensity given by

Consequently, for an absorbing-emitting gray medium with absorption

coefficient a , the total energy emitted by a volume element is

If spectral variations of n are known to be important, then an integration

over wavelength must be included, provided of course that the data for
n as a function of wavelength is known.
From equation (9-13) it might appear that because n > 1, the intensity
radiated from a dielectric medium into air could be larger than the usual
blackbody radiation i;. This is not the case as some of the energy emitted
within the medium is reflected back into the emitting body at the medium-
air interface. Consider a thick dielectric medium (K=O) at uniform tem-
perature and with refractive index n. The maximum intensity received
at an element dA on the interface from all directions within the medium is
n2ii. Only the energy within a cone having a vertex angle P,,,, relative
to the normal of dA will penetrate through the interface; for incidence
angles larger than the energy will be reflected back into the medium.
Hence, the energy received at dA that leaves the medium is

n2i6dA cos P 217 sin P d p = 2.rrn2ikdA sinVP,l,ax


From equation (9-8) with n, = 1 and nz = 11 in this case, sin PJltaZ=

so that the total hemispherical emissive power leaving the interface is

Dividing by rrdA gives i6 as the maximum diffuse intensity leaving the

interface which is the expected blackbody radiation.

EXAMPLE9-1: An elemental volume dV is located at x= 1 cm into

a glass plate from the plate interface with air (as in fig. 9-2). Diffuse-
gray radiation of intensity 10W/(cm2)(sr)in the air is entering the glass
(note this intensity is entering the glass so the reflectivity of the interface
has already been accounted for). If the absorption coefficient of the glass
is 0.005 cm-I and its refractive index is n = 1.75, determine the tempera-
ture at dV as a result of only this incident intensity. Assume the radiation
absorbed from the surrounding glass is small, and neglect heat
An energy balance on dV states that the emitted energy will equal
the incident energy that is transmitted through the glass and absorbed
by dV, that is,

From equation (9-7) the intensity i61,ss (0, P) in the glass at the glass
surface is related to the entering intensity in air ihir (0) by i~,,,, (0, p )
=In2ihi, (0). Since the path length from the glass surface to dV is x/cos P ,
the intensity iGlaSs(x, p ) at dV is given by Bouguer's law as

Substituting into the energy balance and solving for T4 gives

Over part of the 417 solid angle surrounding dV there will be no energy
incident on dV, so that the integration limits on P must be derived by

considering the restrictiorl of equation (9-8). This gives

exp (-5) sin fi dp

Let p = c o s fi to obtain

T -- 277-iLir( 0 )
exp (- 7 ) d p
From equation (2-45) this becomes

T 4 -- 2"i:"r(o)
[ E ( a~ x ) -Io
cos p.,or
exp (-$)d p ]
Now let y = p / c o s P,,ln.r to obtain

T 4= 2ri:,ir(O) ax
Ee (ax) - cos finlcrrE2

Substituting numerical values gives

The E p values can be found from table I1 in the appendix to give

In this example, the effect of the surface reflectivity in determining the

intensity of the radiation that is able to cross the interface and enter
the material was not explicitly treated. It can be introduced for optically
smooth surfaces by using the electromagnetic theory relations for
reflectivity from chapter 4 of volume I.

Gardon (refs. 2 to 5) has treated problems of thermal radiation in glass,

where effects of the refractive index are substantial. Reference 2 in-
cludes some analysis of perpendicular and parallel polarization contri-
butions; in references 3 and 4, a comprehensive analysis of the heat
treatment of glass is given. This analysis of heat treatment includes the

effects of conduction within the glass and convection at the surface. In

reference 5, a review of radiant heat transfer as studied by researchers
in the glass industry is given, and a digest of m ~ l c hof the literature on
the subject up to 1961 is presented. Condon (ref. 6) has given a more
recent review of radiation problems in the glass industry from the view-
point of the astrophysicist.
As mentioned in section 7.3, McConnell (ref. 7) has studied the radia-
tion effects on a space vehicle that has a layer of frost deposited on its
surface. An external source of radiation (the Sun or a diffuse source) is
assumed to radiate to the frost layer, and the resulting temperature
profile in the frost is analyzed as modified by sublimation of the frost at
its free surface. The refractive index of the frost must be accounted for
because of the modification of the usual radiation equations as outlined
in example 9-1.

9.3.3 Radiative Transfer Between Dielectrics Spaced Closely Together

A highly effective insulation can be constructed from many layers of
radiation reflecting films separated by vacuum to provide a series of
alternate radiation and conduction barriers. One construction is to
deposit highly reflecting metallic films on both sides of thin sheets of
plastic. The sheets are then spaced apart by placing between them a
cloth net with large open area between the fibers. Typically a stacking
of 50 radiation shields per inch of thickness can be obtained in this
manner. An important use of multilayer insulation is in low temperature
applications such as insulation of cryogenic storage tanks.
The multilayer insulation can be quite effective. A conventional
analysis of radiation between surfaces in vacuum shows that if two gray
parallel plates with surface emissivity 6 are separated by one radiation
shield also having emissivity 6 on both sides, the heat radiated between
the plates will be reduced to one-half its value without the shield. The
use of n shields all of emissivity e will reduce the heat flow to l / ( n 1 )
of the uninsulated value. The additional effect to be discussed here arises
because the reflecting layers are spaced very close to each other; the
question is whether such small spacings can have any influence on the
radiative transfer.
The situation of transfer between closely spaced surfaces was ex-
amined by Cravalho, Tien, and Caren (ref. 8) who considered the geome-
try shown in figure 9-3. The geometry consists of two semi-infinite
dielectric media having refractive indices nl and n 3 , separated by a
vacuum gap. In the usual analysis for radiative transfer between two
surfaces (e.g., 1 and 3), as given by equation (5-11) of volume 11, the
heat flux transferred across the gap is given by

I PI\ Vacuum

FIGURE 9-3.-Reflection and transmission of electromagnetic wave in gap between two

dielectrics: Tn < T I .

and the spacing between the plates does not appear. When the spacing
between the surfaces is very small, however, there are two effects that
enter which are a function of spacing. The first effect is wave inter-
ference, in which a wave reflecting back and forth in a gap between two
dielectrics may undergo cancellation or reinforcement.

T h e second effect is radiation tunneling. Figure 9-3 reveals that for

ordinary behavior at an interface as discussed in section 9.3.2 some of
the radiation in medium 1 traveling toward region 2 can undergo total
internal reflection at the interface when nl > nz. For ordinary radiative
behavior this would occur when the incidence angle P1 i s equal to
or larger than Brewster's angle given by equation (9-12), that is,
PI 2 sin-'(n~/nl).When region 2 in figure 9-3 is sufficiently thin, however,
electromagnetic theory predicts that, even for an intensity incident at
pl greater than Brewster's angle total internal reflection will not occur.
Rather, part of the incident intensity will propagate across the thin
region 2 and enter medium 3. This effect is radiation tunneling as viewed
As shown in reference 8, both tunneling and interference can become
important only when the spacing between radiating bodies separated
by vacuum is less than about A,,,, aac(T3)which is the wavelength in
vacuum at maximum blackbody emissive power from a surface at the
sink temperature T3. The A,,,,,, ,,,(T:l) is found from Wien's displacement
law as C3/T:3 (eq. (2-17) of vol. I). The tunneling and interference effects
also depend on temperature. Even for very small spacings on the order

FIGURE9-4.-Effects of wave interference and radiation tunneling on radiative transfer

between two dielectric surfaces; n 1= na = 1.25; A,,,,,(T3) = 0.28978/n3T:i cm. (From
ref. 8.)

of Alllnz,vc,c(7'~),the effects become very small at normal temperatures,

and hence are only important in certain cryogenic applications where
temperatures of a few degrees absolute are encountered. Figure 9-4
shows some representative results under conditions giving maximum
effects and illustrates the influence of temperature T I .Note that All1,,(T3)
in figure 9-4 is the wavelength in medium 3 and hence is given by
C3/n3T3 from equation (2-33) of volume I . The conventional solution
referred to in the figure is obtained when wave interference and radiation
tunneling are neglected in the analysis.


Under certain conditions, gases emit much more radiation in the visible
region of the spectrum than would be expected from the absorption co-
efficients of the gas species that are present in chemical equilibrium.
For example, the typical almost transparent blue flame of a bunsen
burner can be made into a smoky yellow-orange flame by changing only
the fuel-air ratio. Such luminous emission is usually ascribed to hot

carbon (soot) particles that are formed because of incomplete combustion

in hydrocarbon flames. There is room for argument even here. Echigo,
Nishiwaki, and Hirata (ref. 9) and others have advanced the hypothesis,
supported by some experimental facts, that the luminous emission from
some flames is due to the emission from vibration-rotation bands of
chemical species that appear during the combustion process prior to the
formation of soot particles. Since the formation of soot is the most
widely accepted view, the radiation from soot will be emphasized here
when discussing luminous flames.
Combustion in the general case is a very complicated chemical process
often consisting of a system of chemical reactions occurring in series and
parallel. The combustion process involves a variety of intermediate
chemical species. The composition and concentration of these inter-
mediate species cannot be predicted very well unless complete knowl-
edge is available of the reaction kinetics of the flame, and this knowledge
will not usually be at hand. Because the radiation properties of the flame
depend on the distributions of species and the temperature variations
within the flame, a detailed prediction of radiation from flames is not
often possible by knowing only the original combustible constituents and
the flame geometry. Because of these difficulties, it is usually necessary
to resort to empirical methods for predicting radiation from systems
involving combustion.
In order to facilitate the present discussion, let us separately examine
two facets of predicting radiation from flames: (1)The calculation of a
theoretical flame temperature by considering the chemical energy re-
lease and without accounting for heat loss by radiation, and (2) the more
complex problem of radiation from a gas containing solid particles which
will alter the theoretical flame temperature.

9.4.1 Theoretical Flame Temperature

To present empirical correlations of radiation from flames, a charac-
teristic parameter is the average temperature of a well-mixed flame as a
result of the addition of chemical energy. Fortunately, well-developed
methods exist (refs. 10 to 12) for computing the theoretical flame tempera-
ture of a given combustion system from available thermodynamic data.
The effect of preheating either the fuel or oxidizer or both can be in-
cluded. Such calculations assume complete combustion of the fuel and
no heat losses. The process of flame temperature computation can be
conveniently shown by an example.

EXAMPLE9-2: Using the mean heat capacity data of figure 9-5

(adapted from ref. 10) and the heat of con~bustionfrom table 9-11, cal-
culate the theoretical temperature of an ethane flame burning with 100

400 800 1200 1600 MOO 2400 2800
Temperature, T, K

FIGURE9-5.-Mean heat capacity of various gases averaged between T and 298 K.

Maximum flame temperature, K
(combustion with dry air at 298 K)
Heat of
Fuel combustion, Theoretical Theoretical
J/kg (complete (with dissocia- Experimental
con~bustion) tion and

Carbon monoxide (CO) 4.83 X 10' 2615 ..........................................

Hydrogen (HI) 12.0 2490 ..........................................
Methane (CHI) 5.0 2285 2191 2158
Ethane (C2H6) 4.74 2338 2222 2173
Propane (C3Hx) 4.64 2629 2240 2203
n-Butane (CIHlo) 4.56 2357 2246 2178
n-Pentane (C5HI2) 4.53 2360 ......................................
Ethylene (C2H4)
Propylene (C&)
Butylene (CIH8)
Amylene (C5H10) 2477
Acetylene (C2HI) 2859
Benzene (CsHa) 2484
Toluene (C,;H5CH,) 4.09 2460

percent excess air (by volume). The ethane is supplied at room tempera-
ture (25O C) and the air feed is preheated to 500' C. The flame is burning
in an environment at a pressure of 1 atm.
The theoretical flame temperature T is computed by using energy con-
servation and assuming no heat losses. The energy in the combustion
constituents plus the energy of combustion is equated to the energy of
the combustion products. This gives

(energy in feed air and) , (energy released)

T - Tref=
\ fuel above Tref )'\ by combustion )
[total mass of), [mean heat capacity)
\ products ) ^ \ of products )
For ethane, assuming complete combustion, the reaction is

Let us assume that 2 kilogram-moles of ethane are burned. Then 7 moles

of oxygen are consumed during combustion. Since there is 100 percent
excess air, only one-half of the feed air contributes oxygen to the combus-
tion process, so a total of 14 moles of oxygen are introduced in the feed
air. Oxygen makes up 21 percent by volume of the feed air, and since
mole fraction is equal to part by volume, the moles of air used are

Thus, the total input to the combustion process consists of 2 moles of

ethane fuel and 66.67 moles of air. The sensible heat in the feed com-
ponents above a reference temperature is

where C p is the mean heat capacity between the reference temperature

and the input temperature Ti. Using data from figure 9-5, which has a
reference temperature of 298 K (25' C), gives

Hi='[mcI)(Ti-Tref)let~zn,,e+ [mc1)(Ti-- Tref)]a i r

= O + 66.67 X 2.93 X 104(773-298) = 9.28 X los joules (J)

where the ethane contributes nothing as it is supplied at Tref. The heat


released by combustion BH is found by using the heat of combustion

from table 9-11 and the fact that the molecular weight of ethane is 30;
that is,
AH=2 moles x 3 0 - kg X 4.74 x 107 -= 28.4 X l o RJ
mole kg

The numerator of equation (9-15) is then

Nitrogen, which makes up about 79 percent by volume of air, remains

from that portion of the feed air that supplied the oxygen for combustion.
The amount of nitrogen in the combustion products is then

33.3 X 0.79 = 26.3 moles

The total quantity of products after co~nbustionis (in moles)

Carbon dioxide (C02) . . . . . . . . . . . . . . . . . . . . . . . . . 4
Water vapor (H20) . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Air . . . . . . . . . . . . . . . . . . . . . . . . . . 33.3 (half of feed air)
Nitrogen (Nz) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26.3
To find the denominator of equation (9-15), the individual quantities
are summed

However, c,depends on the product temperature T, which is the flame

temperature and is not yet known. We must estimate T to determine
the cp, values and then substitute the quantities into equation (9-15).
If the calculated flame temperature agrees with the assumed value, the
solution is finished. Otherwise, a new tenlperature is estimated, H p r o d
is recalculated and substituted into equation (9-15), and this procedure
is continued until the assumed andicalculated flame temperatures agree.
A table of calculations is as follows:

Assumed Mean heat capacity, F,,, J/(kg)(mole)(K) Product Calcu-

flame entlialpy, lated
tempera- H,,,,,,i, J tempera-
ture, K C01 Air ture, K


Because 2 400 K change in the assumed flame temperature produced

only about a 20 K change in the calculated flame temperature, a value
of 1853 K is estimated as within a few degrees of the converged result.

In this example, it was assumed that the combustion process is com-

plete and that no dissociations of the combustion products occur. In
addition, no consideration has been given to energy loss from the flame
by radiation, which would lower the flame temperature. Methods for
including these effects are discussed in reference 12. A list of theoretical
flame temperatures (no radiation included) is shown in table 9-11 for
various hydrocarbon flames. Results for complete combustion with dry
air are shown, followed by calculated results modified to allow for dis-
sociation and ionization of products. The latter are compared with experi-
mental results. In addition, the heats of combustion of the substances
are shown. All data are from references 12 and 13. Extensive tabulations
of similar data for over 200 hydrocarbons are given in references 10 and
Let us proceed to a consideration of the radiation emitted by a non-
luminous flame now that its average temperature is known.

9.4.2 Radiation From Nonluminous Flames

The phenomena involved in radiation from the nonluminous portion
of the combustion products are fairly well understood. The complexities
of the chemical reaction are not too important here since it is the gaseous
end products situated above the active burning region that are being
considered. During combustion, chemical potential energy is released
by the reaction of the fuel and oxidant atoms. This results in radiation
in spectral lines and bands produced by the various types of transitions
between energy states. In most instances a hydrocarbon combustion is
being considered and the radiation is from the COP and H z 0 bands in
the infrared. For flames a few or more feet thick as in commercial fur-
naces, the emission leaving the flame within the COz and H z 0 vibration-
rotation bands can be close to blackbody emission.
The gaseous radiation properties and methods of chapter 5 can be
used to compute the radiative heat transfer from the flame. The analysis
is greatly simplified if the gas is well mixed so that it can be assumed
isothermal. For a nonisothermal condition the gas can be divided into
approximately isothermal zones, and the convection within the gas can
also be included if the circulation pattern within the combustion chamber
is known. A nonisothermal analysis with convection was carried out in
reference 14 for cylindrical flames.

EXAMPLE9-3: In example 9-2, the combustion products were 4 moles

of Con, 6 moles of Hz0 vapor, 33.3 moles of air, and 26.3 moles of N2.
Assume these gaseous products are in a cylindrical region 4 ft high and
2 ft in diameter and are uniformly mixed at the theoretical flame tem-
perature 1853 K. The pressure is 1 atm. Compute the radiation leaving
the gaseous region using the methods in sections 5.5 and 5.6.
The partial pressure of each constituent is equal to its mole fraction
of the mixture. Then for the Con and HpO the partial pressures are

PC,, = (A) (1 atm) = 0.0574 atm

I)H,~= (A) (1 atm) = 0.0861 atrn

The mean beam length of the gas for negligible self-absorption can be
computed from equation (5-51) as

To include self-absorption, a correction factor of 0.9 is applied so the

mean beam length becomes

Le= 0.9(1.6) = 1.44 ft


Using the gas emittance charts (figs. 5-11 to 5-15) at the flame tem-
perature (3340 OR) results in


eH20=0.029 X 1.08=0.031

The 1.08 factor in EHzo is a correction for the partial pressure of the water

vapor not being zero. In addition, there is a negative correction resulting

from spectral overlap of the C 0 2 and H 2 0 radiation bands. This is
obtained from figure 5-15 at the values of the parameters

pcozL, +pHZ0Le=0.0825 + 0.124 = 0.207 (atm)(ft)

The correction is A€ = 0.002. Then the gas emittance is

The radiation from the gas region is then computed as

Q = e,JaT;= 0.068(10~)0.173X 10-"3340)"= 0.46 X lo6 -

9.4.3 Radiation From and Through Luminous Flames

In the region of the flame that is actively burning, there are several
factors that complicate the radiative transfer. The simultaneous pro-
duction and loss of energy produces a temperature variation and thus
a variation of properties and emission within the flame. The intermediate
combustion products resulting from the complex reaction chemistry
can significantly alter the radiation characteristics from those of the
final products. Soot is the most important radiating product formed when
burning hydrocarbons. The soot emits in a continuous spectrum in the
visible and infrared regions, and as a result of the visible radiation the
flame is called luminous. The soot is quite important as it can often
double or triple the heat that would be radiated by the gaseous products
alone. A method for increasing the flame emission if desired is to promote
slow initial mixing of the oxygen with the fuel so that large amounts of
soot will form at the base of the flame.
Determining the effect soot has on the flame radiation resolves into
two requirements. One of these is to somehow obtain the soot distribu-
tion in the flame. This depends on the type of fuel, the mixing of fuel
and oxidant, and the flame temperature. The soot distribution is too
complicated to calculate from basic principles, so some experimental
knowledge of a given combustion system is needed. The second re-
quirement is to know the radiative properties of the soot. Then if the
soot concentration and distribution are known, a radiation computation
can be attempted. At present the radiant properties of soot are only
known to a first approximation.

If the flames found in both the laboratory and in industry are inciuded,
the individual soot particles produced in hydrocarbon flames genera!y
range in size from a diameter of 50 angstroms (A) to greater than 3000 A.
The soot can be in the form of spherical particles, agglomerated masses,
or is sometimes in long filaments. The experimental determination of
the physical form of the soot is most difficult, as any type of probe that
is used to gather the soot for photomicrographic analysis may cause
agglomeration of particles or otherwise alter the soot characteristics.
The nucleation and growth of the soot particles is not well understood.
Some of the soot can be nucleated in less than a millisecond after the
fuel enters the flame, and the rate at which soot continues to form does
not seem influenced much by the residence time of the fuel in the flame.
It is an unknown precipitation n~echanisnlthat must govern the soot
Stull and Plass (ref. 15) and Siddall and McGrath (ref. 16)have computed
the spectral emittance of luminous flames as a function of the volume
fraction of soot particles present. This was done by the use of Mie theory
(section 8.4.6), which is a direct application of electromagnetic theory,
to obtain the radiation characteristics of the assumed spherical soot par-
ticles (the Mie result will be given in equation (9-22)). The calculations
were carried out using optical properties of a baked electrode carbon at
2250 K as the assumed n and K for soot carbon. The results should b e
valid over a range of temperatures because Howarth, Foster, and Thring
(ref. 17) have shown that the absorption coefficient for carbon particles is
at most a weak function of temperature.
For a beam of radiation passing through a gas containing suspended
soot particles, it has been found experimentally that the attenuation
obeys Bouguer's law, that is,

ii ( S )= ii (0) exp (- AS)

For small particles such that TDIA < 0.25 (where D is particle diameter),
the Mie theory gives in equation (8-30) that the scattering cross section
depends on ( ~ D l h )The
~ . Mie theory also shows that, for the same con-
ditions, the absorption cross section depends on TDIA to the first power
(see eq. (9-22)). Thus the scattering is negligible compared with absorp-
tion, and ah in equation (9-16) is actually the absorption coefficient rather
than the extinction coefficient of equation (1-13). Then as a consequence
of equation (5-44), the spectral emittance of an isothermal luminous gas
volume is written as
€ A = 1- exp (- ahL,) (9-17)

where L, is the mean beam length for the volume. Experimental correlation of soot spectral absorption. - A rela-
tively simple empirical relation for ah that has been found experimentally
in some instances has the form

where C is the soot volume concentration (average volume of particles

per unit volume of cloud) and k is a constant. The A will always be in
microns for the numerical results given here. Hottel (ref. 18) recommends
the use of

i n the infrared region down to X=0.8 pnL. In some more recent experi-
ments, Siddall and McGrath (ref. 16) also found the functional relation of
equation (9-19) to hold approximately. They give in the range from A= 1
to 7 ,urn the following mean values of a:

Source of soot Mean cu for

A=l to7pm

Amy1 acetate
Avtur kerosene
Furnace samples

Thus the 0.95 exponent recommended by Hottel appears reasonable.

In reference 16 the data were also inspected in more detail to see if a
had a functional variation with A that would provide a more accurate
correlation than using a constant a. In some instances a took the form

where a and b are positive constants. Examples are shown in figure 9-6.
In other cases, as in figure 9-7, a more general polynomial was required
to express a as a function of A. Thus, as a generalization of equation
(9-19), for the infrared region
a),=- Ad),)

Amyl acetate soot (two


1 different samples)
Benzene soot
A Petrotherm soot

1 2 4 6 8
Wavelength, h, p m

FIGURE9-6.-Experimental values of a plotted against A for cases where a varies ap-

proximately linearly in 111 A (ref. 16).

0Avtur kerosene soot

0 Benzene soot

Wavelength, A, p m

FIGURE9-7. -Experiniental values of a as a function of A where oc does not vary linearly

with In A (ref. 16).

and letting a be a constant is only an approximation.

In the visible range an inspection by Hottel (ref. 18) of experimental
data led to the recommended form

for the.wavelength region around A = 0.6 p m (say A -- 0.3 to 0.8 pm).

SPECIALIZED EFFECTS IN RADIATING MEDIA 335 Electromagnetic theory prediction of soot spectrul absorp-

tion.-To try to understand the absorption coefficient of a soot cloud
from a more fundamental basis, electromagnetic theory can be employed
(refs. 15, 16, 19, and 20). The absorption coefficient is written as

The product EAA is the spectral absorption cross section, defined in

the same manner as the scattering cross section SAused in equation (8-5).
The N is the number of particles per unit volume, and A is the pro-
jected area of a particle ( A = 7rQ2/4, as particles are assumed spherical).
The Ex by itself is the spectral absorption efficiency factor which is the
ratio of the spectral absorption cross section to the actual physical cross
section of the particle. For the limit of small particles the Mie equations
give the EAfor a small absorbing sphere as

2hD n,K
E),= ---
A [(n2- K2) 2]2+ 4n2~2

where n and K are the simple index of refraction and the extinction
coefficient of the sphere material when the complex index of refraction
is expressed in the form r i = n - i ~ Since
. the optical quantities n and
K are functions of A, equation (9-22) can be written as

Then from equation (9-21)

where C= N7rDR/6is the volume of the particles per unit volume of the
cloud. The ratio

is then a function of wavelength and can be evaluated if the optical prop-

erties of soot are known as a function of A.

In reference 20 the optical properties n and K of propane soot were

measured by collecting the soot and then compressing it on a brass plate.

The values obtained were as follows:

Simple refrac- Extinction

tive index, coefficient,
I1 K


By using these values the ai/C was evaluated from equation (9-25) yield-
ing figure 9-8. Although the ah/C decreases with A as expected from the
form of equation (9-18), it is evident that an approximate curve fit by
straight lines (on the logarithmic plot) would yield exponents on A s o m e -
what different than those of equations (9-19) and (9-20).
The predicted form by Mie theory of a(A) as used in equation (9-19a)
can now be examined in more detail in the infrared region by e q u a t i n g
the expressions in equations (9-19a) and (9-24). This gives

Wavelength, X, pm

FIGURE9-8. -Spectral absorption coefficient divided by v o l u ~ n econce~ltrationfor p r o P

soot (ref. 20).

By evaluating this relation at A= 1, the constant kl applicable to the infra-

red region is found as
kl= 36rF ( 1 ) (9-27)

Taking the logarithm of both sides and solving for a ( A ) give

The optical properties of the soot can then be used in F ( 1 ) a n d F ( 1 ) as

defined in equation (9-25), and a ( A ) can be found. This was done in
reference 16 using the properties of a baked electrode carbon a t 2250 K.
The results are shown in figure 9-9. The trend is the same as t h e experi-
mental curves of figure 9-6, but the a values are larger than t h e experi-
mental values. They are also larger than the average value recommended

Wavelength, A, prn

FIGURE 9-9. -Calculated variation of a with wavelength using properties of b a k e d electrode

carbon at 2250 K (ref. 16).

Similarly by using a green filter

where the green wavelength A, is 0.555 pm.

Now as a simple approximation, equation (9-20) is used for ax in the
visible region. Then E A from equation (9-17) is

where S is the path length sighted through the flame. Substitute equation
(9-35) into equations (9-33) and (9-34) to obtain

These two equations are solved for Tfand Ckz thereby yielding the needed
measure of the soot concentration as well as the flame temperature.
AS an approximation, the Ckz is assumed independent of wavelength
and is used in equations (9-17), (9-19), and (9-20) to yield f o r a path
length S

A. = 1- exp (-g) infrared, A > 0.8 p m (9-39)

Then with these spectral emittances the definition in equation (9-29) can
be used to evaluate the total emittance of the flame as

Some convenient graphs for use in this procedure are given in reference
18. The hope is that the Ckz obtained in this way can be applied to "simi-

lar" flames. This is a very rough approximation; there are so many vari-
ables affecting the flow and mixing in the flame that it is difficult to know
when the flames will have a bimilar character. The detailed nature of
flames is a continuing area of active research.

9.4.4 Roldiatinn From Gases Containing Lurnineus Particles

In addition to the work on hydrocarbon luminous flames in the field of
combustion, other fields involve consideration of radiation from luminous
gases. A common example is the luminosity in the exhaust plume of
solid fueled and some liquid fueled rockets. For a solid fuel the luminosity
may be caused by particles of metal that are added to promote combus-
tion stability. The metal particles are heated to high temperatures and
may undergo oxidation, thereby becoming luminous.
The presence of particles in an otherwise weakly absorbing medium
can cause the mixture to be strongly absorbing. "Seeding" of a gas with
particles, such as finely divided carbon, has been proposed in order to
increase the gas absorption (ref. 27), or as a means of shielding a surface
from incident radiation (ref. 28). These techniques have possible applica-
tion in connection with advanced propulsion systems.
Another use for seeding is in the direct determination of flame tem-
peratures for a nonluminous flame by the line reversal techn.iqne. In this
method, a seeding material such as a sodium or cadmium salt is intro-
duced into an otherwise transparent flame. These materials produce a
strong line in the visible spectrum because of an electronic transition.
The cadmium gives a red line and the sodium a bright yellow line. A con-
tinuous source such as a tungsten lamp is placed so that it may be viewed
through the seeded flame with a spectroscope. The intensity seen in the
spectroscope at the line wavelength is, from the integrated equation of
transfer, equation (2-lo),

If the flame is assumed isothermal and of diameter D and no attenuation

occurs along the remainder of the ~ a t hbetween the continuous source
and the spectroscope, equation (9-41) becomes


In the wavelength region adjacent to the ansorbing and emitting spec-

tral line, the flame is essentially transparent so the background radiation
observed adjacent to the line is il\,co,1r.601crce.Hence, by subtractin::
ii,co,,o.ce from equation (9-42) the line intensity relative to the adjacent
background is

If the flame is at a higher temperature than the continuous source, equa-

tion (9-43) shows that the line intensity in the spectroscope will b e greater
than the continuous background intensity. The line will then a p p e a r as a
bright line imposed upon aless bright continuous spectrum in t h e spectro-
scope. By increasing the temperature of the continuous source, t h e source
term will ovkrride. The line then appears as a dark line on a brighter con-
tinuous spectrum. If the continuous source is a blackbody a n d its tern-
perature is made equal to the flame temperature, then

and equation (9-43) reduces to ii = ii ,,,,, The line will then dis-
appear into the continuum in the spectroscope. This is b e c a u s e the
absorption by the flame and the flame emission exactly compensate. If
the continuous source is a tungsten lamp, the source temperature meas-
urement is usually made with an optical pyrometer.
It is noted that in the derivation of equation (9-43) it w a s assumed
that the flame is transparent except within the spectral line produced
by the cadmium or sodium seeding. If soot isrin the flame, t h e soot
particles absorb, emit, and scatter radiation in a continuous spectrum
along the path of the incident beam. The line reversal t e c h n i q u e is of
less ~ r a c t i c a lutility in this instance as it then depends o n the soot
behavior. The effect of soot is analyzed in reference 29.
Another instance of radiation attenuation by means of p a r t i c l e s is
found in the effect of dust or so-called "grains" that are b e l i e v e d to
exist in the interstellar space and cause reductions in t h e observed
intensity of radiation from stars (refs. 30 and 31).

The phenomenon of luminescence in its various forms is a fairly
common one. The name covers a broad range of mechanisms t h a t result

in emission of radiant energy by the transition of electrons from an

excited state to a lower energy state, where the original excitation took
place by means other than thermal agitation. This is a n example of a
process that is not in local thermodynamic equilibrium (section 1.8).
Because the electronic transitions are between discrete energy states,
the span of wavelengths over which the emission occurs is quite small.
Luminescence, therefore, does not add significant energy to the spectrum
of emission in engineering situations and can almost invariably be
neglected in engineering calculations. However, there are some situa-
tions in which effects other than total energy transport are of interest.
For that reason a very brief mention of luminescence is included here.
Luminescence is categorized in various ways. A common classification
is by duration of the effect. Luminescence that persists over a relatively.
long l o period is called "phosphorescence," a word derived from the
luminescence of white phosphorus.ll Luminescence that persists only
during the influence of some external exciting agent such as an ultra-
violet lamp is called "fluorescence," a name arising from the strong
luminescence shown by fluorspar when so irradiated.
Another categorization is by description of the excitation agent. Thus,
luminescence arising from a chemical reaction such as the oxidation of
white phosphorus is called "chemiluminescence"; luminescence caused
by a beam of incident electrons as on a color TV screen is "cathodolu-
minescence"; a biochemical reaction producing luminosity, as in fireflies
and some marine animals, is called "bioluminescence"; luminous
emission by the presence of an electric field as in certain commercial
panel lamps is "electroluminescence"; and luminescence due to photon
bombardment is often called "photoluminescence." The latter effect is
caused by the same mechanism that causes the laser to function. Other
mechanisms can cause luminescence in materials, but descriptive
terms have not yet been coined. Examples are proton bombardment,
which is believed to be responsible for the luminous red patches observed
on the surface of the Moon, and nuclear reactions that cause luminous
emission of radiation.
Because luminescence is common to materials at room temperature,
it obviously cannot be predicted by the usual laws that govern thermal
radiation as these would predict no visible radiation a t such temper-
atures. This is the origin of the term cold light for fluorescent lamp emis-
sion. Rather, the quantum mechanical properties of such luminescent
materials must be examined to explain their behavior. Some detailed
material on luminescence is contained in references 32 and 33. T h e

"'hfeaning in comparison with relatively short, tu speak precisely.

is named for the Greek word meaning "liglit carrying."
" Phosl~horusitself

calculation of luminescence effects is outside the scope of this work and

will not be treated further here.


The treatment of transient phenomena when radiation is present has
been sparsely treated in the literature. Some problems dealing with the
effects of nuclear weapons and some situations in astrophysics require
inclusion of transient effects.
The equation of transfer as derived in chapter 2 neglected changes in
radiation intensity with time. The equation of transfer is written for a
beam of radiation of intensity i; traveling in the S direction. As the radia-
tion travels through the differential length from S to S dS,its intensity
is increased by emission and decreased by absorption. Also during the
residence of the radiation within dS, the intensity can change with time.
The residence time is dt= &/c where c is the speed of propagation in
the medium. Hence, the change in i; can be written as

By substituting for di;, the equation of transfer (2-4) becomes

1 aiL(t,S)
c at
+ d i ;as
= a ~ ( tS, ) [iLb(t,S ) -i; ( t , S) ] (9-44)

Since the conditions such as temperature within the medium a r e chang-

ing with time, the absorption coefficient is a function of t i m e a s well as
Because the speed of light is usually very large c o m p a r e d with the
other quantities in the transient term, the transient t e r m i s usually
very small, and the equation of transfer reverts to the steady-state form
that has been given throughout this work. In some a n a l y s e s directed
towards the study of nuclear weapons (refs. 17 and 18 of c h a p t e r 6), the
transient term is included. To better understand the t r a n s i e n t term,
consider as a simple illustration what the radiative behavior w o u l d be if
a thick uniform medium at temperature T1 instantaneousf-Y had its
temperature increased to a higher uniform value T2. The m e d i u m would
then be at T2 but the intensity within the medium would h a v e to change
from i;, ( T I ) to i;, ( T 2 ) .During this process, the radiation w o u l d not be
in equilibrium. The equation of transfer reduces to ( a s s u m i n g as an aP-
proximation that ax can be used in the emission term, w h i c h is an
equilibrium assumption),

After integrating with the condition i,(=i,(,,(T,) at t = O , the result is

The radiation relaxation time (time to change by a factor of e) for equilib-

rium to he reestablished is thus l / c n x ( T 2 )which is usually very short
for reasonable values of ax in view of the large value of the propagation
velocity c in the medium.
In the preceding illustration, it was assumed that the medium tem-
perature could be instantaneously raised so that at the beginning of the
transient the radiation intensity was not in equilibrium at the black
radiation value corresponding to T2. Generally the temperature change
of a medium would be governed by the heat capacity of the medium,
and consequently transient temperature changes would be much slower
than the radiation relaxation time. Hence, when coupled with the tran-
sient energy conservation equation, which contains a heat capacity
term, the unsteady term in the equation of transfer would be negligible.
This is why the steady form of the equation of transfer, as derived in
chapter 2, can be instantaneously applied during almost all transient
heat transfer processes.

EXAMPLE9-4: A gray medium is in a slab configuration originally

at a uniform temperature To. The absorption coefficient is a, and t h e
slab half thickness is D. The heat capacity of the medium at constant
volume is c, and its density is p. At time t=O, the slab is placed in
surroundings at zero temperature. Neglecting conduction and convec-
tion, discuss the solutions for the temperature profiles for radiative
cooling when a is very large and when a is very small.
At the slab center which is located at x = 0, the condition of symmetry
provides the relation for any time,

At time t = 0 there is the condition

T=T,; t=O,x

As discussed in section 2.6.2 for radiation only being included. there

rvill be a temperature slip at the boundaries x=-+D, so that the
perature at the boundaries will be finite rather than being equal to the
zero outside temperature. If heat conduction were present, the
perature slip would not exist.
For a large a the diffusion approximation can be employed and
equation (3-25) the heat flux in the x direction is

By conservation of energy

Combining these two equations to eliminate q gives the transient energy

diffusion equation for the temperature distribution in the s l a b

Defining dimensionless variables as follows:


The initial condition and the boundary condition at .xs become,


7(0, t * ) =O
c) K

At the boundary K = ccD, a slip condition must be used. Usir+g equation

(3-45), when the surroundings are empty space at zero t e m p e r a t u r e , the
et,,0=0, and the e l c = l , SO that at the exposed boundary of medium
for any time

Similar relations apply at x=- D. For these conditions solution by nu-

merical techniques is probably necessary.
For a small absorption coefficient, and since there are no enclosing
radiating boundaries present, the emission approximation (section
3.3.2) can be applied. For very small a the medium is optically so thin
that it is at uniform temperature throughout its thickness at any instant.
From the results of example 3-2, the heat flux emerging from each
boundary of the layer is
q= 4aaT4D
then becomes
pc, -=- 4aaT4
or, in dinlensionless terms,


0 .2 .4 .6 .8 1.0
Optical depth, K / K ( X= D)

FIGCJRE9-12.-Dimensionless tenlperature profiles as a function of time for radiative

cooling of a pray slab; optical thickness K ( Z = D )= 1.0. (Fro111 ref. 34.)

I n t e g r a t i n g w i t h t h e c o n d i t i o n t h a t O = 1 a t t"= 0, t h e t r a n s i e n t t e m p e r -
a t u r e t h r o u g h o u t t h e slab is t h e n g i v e n by

Viskanta and Bathla (ref. 34) h a v e obtained numerical s o l u t i o l l s t o the

t r a n s i e n t form of the complete e q u a t i o n of t r a n s f e r , a l o n g with t h e
limiting s o l u t i o n s derived here. Some of t h e i r r e s u l t s for intermediate
o p t i c a l thickness are shown in figure 9-12. Numerical s o l u t i o n s f o r
spherical g e o m e t r i e s are found in references 35 and 36. S o l u t i o n s to m o r e
i n v o l v e d t r a n s i e n t problems by use of M o n t e C a r l o are mentioned in
s e c t i o n 6.7.

1. HODGMAN, CHARLESD., ED.: Handbook of Chemistry and Physics. 38th ed., Chemical
Rubber Publ. Co., 1956-1957.
2. GARDON, ROBERT:The Emissivity of Transparent Materials. J. Am. Cer. Sot., vol. 39,
no. 8, Aug. 1956, pp. 278-287.
3. GARDON, ROBERT:Calculation of Temperature Distributions in Glass Plates Under-
going Heat-Treatment. J. Am. Cer. Soc., vol. 41, no. 6, June 1958, pp. 200-209.
4. GARDON,ROBERT: Appendix to Calculation of Temperature Distributions ill Glass
Plates Undergoing Heat Treatment (ref. 3). Mellon Inst., Pittsburgh, 1958.
5. GARDON,ROBERT:'AReview of Radiant Heat Transfer in Glass. J. Am. Cer. Sot., vol.
44, no. 7,July 1961, pp. 305-312.
6. CONDON, EDWARD U.: Radiative Transport in Hot Glass. J. Quant. S p e c t r o s c Radiat.
Transfer, vol. 8, no. 1, 1968, pp. 369-385.
7. MCCONNELL, DUDLEY G.: Radiant Energy Transport in Cryogenic Condensates. Paper
presented at Enpineering Colloquium, Kentucky University, Lexington, K y . , Nov. 30,
8 . CRAVAI~HO, E. G.; TIEN,C . L.; AND CAREN,R. P.: Effect of Small Spacings or1 Radiative
Transfer Between Two Dielectrics. J. Heat Transfer, vol. 89, no. it, Nov. 1967, pP.
9. ECHIGO,R.; NISHIWAKI,N.; AND HIRATA,M.: A Study on the Radiation o f Luminous
Flames. Eleventh Symposium (International) On Combustion. The ~ o n l b u s t i o l l
Institute, 1967, pp. 381-389.
Engineers' Handbook. Fourth ed., McGraw-Hill Book Co., Inc., 1963.
Energy Balance. Vol. 1 of Chemical Process Principles. Second ed., J o h n Wiley &
Sons, Inc., 1954.
12. GAYDON,A. G.; AND WOLFHARD,.H. G.: Fla~iles,Their Structure, Radiation, and
Temperature. Second ed., Mac~iiillanand Co., 1960.
13. BARNETT.HENRYC.; AND HIBBARD.ROBERTR.. EDS.: Basic Considerations in the
Combustion of Hydrocarbon Fuels with Air. NACA Rep. 1300,1957.
14. HOTTEL,H. C.; AND SAROFIM,A. F.: The Effect of Gas Flow Patterns o m Radiative
Transfer in Cylindrical Furnaces. Int. J. Heat Mass Transfer, vol. 8, no. 8, Aug. 1965,
pp. 1153-1169.


. N.: Emissivity of Dispersed Carbon Particles.
J . Opt. Soc. Arn., vol. 50, no. 2, Feb. 1960, pp. 121-129.
16. SIDDALL,R. G.; ANI) ~ I C G R A T H1., A,: The Eilrissiviiy of Luminous Flames. Ninth
Symposium (International) on C:ombustion. W. G. Berl, rtl., Academic l'ress, 1963,
pp. 102-110.
17. HOWARTH,C. R.; FOSTER.P. J.; AVD THRINC;.M. W.: Ttie Effect of Teniperature 011
the Extinction of Radiation by Soot Particles. Tliircl International Heat Transfer
Confereilce. Vol. 5, AIChE, 1966, pp. 122-128.
18. HOTTEL,H. C.: Radiant Heat Transmission. Heat Transmission. Thirtl ed., \William H.
McAdams, 5IcGraw-Hill Book Co., Inc., 1954, cli. 4.
19. HAWKSLEY,P. G. W.: The bIetliods of Particle Size A~Ieasurenrents.Part 11. Optical
Illethods and Liglit Scattering. Brit. Coal Utilization Res. Assoc. Monthly Bull.,
vol. 16, nos. 4 and 5,1952, pp. 134-209.
20. DALZELL,W. H.; AND SAROFIM.A. F.: 0j)tical Constants of Soot and Their Applica-
tion to Heat-Flux Calculations. J. Heat Transfer, vol. 91, No. I , Feb. 1969, pp. 100-104.
21. ERICKSON,W. D.; WILLIAMS,G. C.; AND HOTTEL, H. C.: Liglit Scattering Meas-
urements on Soot in a Benzene-Air Flame. Combt~stionand Flame, vol. 8, no. 2,
June 1964, pp. 127-132.
22. THRING,M. W.; BEER,J. YI.; AND FOSTER,P. J.: The Radiative Properties of 1,uminous
Flanies. Third International Heat Transfer Conference. Vol. 5, AIChE, 1966, p P .
23. THRING,M. W.; FOSTER,P. J.; MCCRATH,I. A.; AND ASHTON,J . S.: Prediction of t h e
Emissivity of Hydrocarbon Flames. International Developments in Heat T r a n s f e r
ASME, 1963, pp. 796-803.
24. SATO,TAKASHI;AND MATSIJMOTO,KYUICHI:Radiant Heat Transfer from L u m i n o ~ l s
Flame. International Developments in Heat Transfer. ASME, 1963, pp. 804-811-
25. YAM, S.; A N D INOUE,H.: Radiation from Soot Particles in 1,uminous Flames. E i g h t h
Sytnposiuni (International) On Combustion. Williams and Wilkins, 1962, pp. 288-293.
26. BONE, WILLIAMA.; AND TOWNEND, DONAI.DT. A.: Flame antl Comhustioll in G a s e s -
Lo~igmans,Green and Co., Ltd., 1927.
27. IANZO,CHESTERD.: AND RA(;SDAI.E.ROBERT(;.: Heat Transfer to a Sreded F l o w i n %
(;as from an Arc Enclosed by a Quartz Tube. Proceedings of the 1964 Heat T r a n s f e r
and Fluid A31echanics Institute. Warren H. Giedt and Salornon Levy, eds., S t a l l f o r d
Univ. Press, 1964, pp. 226-244.
28. HOWELL,JOHN R.; AND RENKEL,HAROLDE.: Analysis of the Effect of a S e e d e d
Propellant Layer on Thermal Kadiation in the Nozzlr of a Gaseous-Core N u c l e a r
Propulsion System. NASA T N D-3119, 1965.
29. THOMAS,D. LYDDON:Problen~s in Applying the Line Reversal Metlrod of T e n ? .
perature Measurement to Flanies. Combustion and Flame, vol. 12, no. 6, Dec. 1 9 6 8 ,
pp. 541-549.
30. CREENBERG, J. MAYO;A N D ROARK,T. P., EDS.: Interstellar Grains. NASA S P - ~ ~ O >
31. DONN,B.; A N D KRISHNA SWAMY,K. S.: Extinction by Interstellar Grains, Mie I'articles
and Polycyclic Aronratic Molecc~les.Physica, vol. 41, no. 1, Feb. 21,1960, pp. 144-1 5 O .
32. CIJRIE,DANIEI,(G. F. J. (;.ARLICK. TRANS.):1,uminescence in Crystals. John W i l e y &
Soils, Inc., 1963.
33. PRINGSHEIM. PETER: Fluorescence and Phosl>llorescence. Interscience Publ., 194.')
34. VISKANTA,RAYILIONII; A N D BATHLA,PRITAM S.: Unsteady Energy Transfer i n a
Layer of Gray (;as I)y Thermal Radiation. Zeit. f. r1t1y.e~.Math. Pltys., vol. 18. n o - 3.
1967. pp. 353-367.

35. VISKANTA, R.: AND LAI.L, 1'. S.: Transient Cooling of a Splrerical Mass of Hig11-
Temperature Gas by T h e r n ~ a lRadiation. J. Api~1.S'Iech., vol. 32, no. 4, Dec. 1965,
pp. 740-746.
36. VISKANTA,R.; AND I.ALL, 1'. S.: Transient Heating a ~ l t lCooling of a Spherical Mass
of Gray Gas hy Thermal Radiation. Proceedings of the Heat and Transfer Fluid
blechanics Institute. 91. A. Saad and .I. 4. Sliller, eils., Stanford Univ. l'rrss. 1966,
pp. 181-197.
For the convenience of the reader, some of the important constants
used in radiative transfer theory are gathered here in table I.

First Bohr electron radius a,= h"m,e" 0.5292 X cm

Speed of light in vacuum c,=2.9979 X 10"' cmlsec
Electron charge e=4.803 X 10-I" esu
Planck's constant h = 6.625 X lo-" (erg)(sec)
h = h/27r= 1.054 X (erg)(sec)
Boltzmann constant k = 1.3804 X lO-'%rg/K
Electron mass m e = 9.108 X
Classical electron radius r , = e2/m,c; = 2.818 X 10-':I ern
Atomic unit cross section r a p = 0.880 X 10-1" cni2
Titomson cross section o.r= 87rri/3 = 6.652 X 10-2%cm"
Electron volt 1 eV = 1.602 X 10-'"erg
Temperature associated with 1 eV 1 ei7/k=11605K
Rest energy of electron n ~ , c=z 8.186 X lo-' erg
Ionization potential of hydrogen atom e2/2a,= 2.rr'e4m,,/h2= 13.60 cV


A summary of some useful exponential integral relations is presented
here. Additional relations are given in references 1 to 3.
For positive real arguments, the nth exponential integral is defined as

and only positive integral values of n will be considered here. An alter-

nate form is

By differentiating equation (Al) under the integral sign, t h e recur-

rence relation is obtained

Another recurrence relation obtained by integration is

nEll+l( x )= exp ( - x ) -xE,,(x) n 2 1 (A41

Also integration results in

By use of equation (A4), all exponential integrals can be reduced to

the first exponential integral given by

El ( x )=
di p-I exp
( 5 l d p

Alternate forms of El ( x ) are

~,(x) =Ix til exp ( - x t ) d t =

:1 1-1 exp ( - t ) d t

For x=O the exponential integrals are equal to

For large values of x there is the asymptotic expansion

Series expansions are of the forill


where y=0.577216 is Euler's constant. The general series expansion

given in reference.3 is



Tabulations of E , , ( x ) are given in references 2 and 3. An abridged

listing given in table I1 is included here for convenience.


[From ref. 21

E,(X) Es(x)

1.0000 0.5000
.9497 .4903
.9131 .4810
,8817 .4720
.8535 ,4633
.8278 .4549
.8040 .4468
.7818 .4388
.7610 .4311
.7412 .4236
,7225 .4163
,5742 ,3519
,4691 .3000
.3894 .2573
.3266 ,2216
.2762 .I916
.2349 .I661
,2009 .I443
.1724 .I257
.I485 ,1097
.lo35 .0786
,0731 .0567
.0522 ,0412
.0375 .0301
.0272 .0221
.0198 '0163
.0145 .0120
.0106 .0089
.0078 .0066
.0058 .0049

1. CHANDRASEKHAR, Radiative Transfer. Dover Publications, Inc.,
2. VLADIMIR:Basic Methods in Transfer Problems. Dover ~ u b l i c a t i o ~ ~ +
Inc., 1963.
3. MILTON; AND STEGUN,IRENE A., EDS.: Handbook of ~ a t h e r n a f i ~ ~ l
Functions with Formulas, Graphs, and Mathematical Tables. Appl. Math. Ser. 55$
Nat. Bur. Standards, 1964.

Absorptance, 24 Broadening, line,

Absorption, 4 collision, 129, 132
band correlations, 138, 146 Doppler, 129, 131
band overlap, 152, 187 natural, 129, 131
bound-bound, 7, 128 Stark, 129, 134
bound-free, 8, 137 Carbon dioxide radiation, 2
free-free, 8, 137, 138 band, 5,149
line, 7, 16, 128, 134 charts, 184, 185
Absorption coefficient, mixture with water vapor, 186
definition, 13, 15 Carbon monoxide band radiation, 149
effective mean, 96 Channel flows with radiation, 270
incident mean, 46,93 Closely spaced dielectrics, 322
mass, 13 Coefficient,
Planck mean, 45, 66, 67, 93, 224 absorption, 13, 15, 19
Rosseland mean, 78,93 emission, 23
true, 19 extinction, 12
Absorption factors, scattering, 13, 20, 283
geometric mean, 164 Cold medium approximation, 62, 70
geometrical, 164 Collision broadening, 129, 132
Addition of radiation and conduction, 251 Conduction-radiation parameter, 244
Angular frequency, 113 Coupled problems, 241, 264
Attenuation, 12 additive solution, 251
atmospheric, 1 boundary layer, 265
Bouguer's law, 13, 16 channel flows, 270
Band absorptance, 138 diffusion method, 254
Band models, Monte Carlo, 237
Elsasser, 143 radiation and conduction, 237, 241, 243,
exponential wide, 148 246,247,251
statistical, 144 radiation, conduction and convection,
Band width, 237,241,264
correlations, 138, 148 Cross section,
effective, 142 absorption, 335
limits, 196 scatter, 280
tables, 146, 148, 149 Curtis-Godson approximation, 202
Bohr model of atom, 115 d e Broglie, Louis, 117
Boltzmann distribution, 123, 132 Degenerate states, 121
Bouguer, Pierre, 13 Detailed balancing, 123
Bouguer's law, 13, 16 Differential approximation, 102
Boundary layer with radiation, 265 boundary conditions, 106
optically thin layer, 266 equation of transfer, 106
optically thick layer, 268 solutions for simple geometries, table,
Bremsstrablung, 8 109
Brewster's angle, 317 Diffraction from sphere, 292

Diffusion method, 62, 71,74, 254 Exponential integrals,

jump between two absorbing-emitting definition, 56
regions, 82, 86 in equation of transfer, 55, 248
jump boundary condition, 79 relations between, 353
radiation and conduction, 254 table of values, 356
Rosseland diffusion equation, 74, 75, 78 Exponential wide band model, 148, 149
scattering, 308 Extinction coefficient, 12
Diffusion solutions, Flames, 324
concentric cylinders, 89 luminous, 331
concentric spheres, 89, 90 nonluminous, 329
parallel plates, 83 theoretical temperature, 325
table, 89 Flux vector, 58
Direct exchange area, Gaunt factor, 128
gas-gas, 211 Geometric mean beam length, 191, 192,
gas-surface, 208 194
surface-gas, 210
Gray gas, 83, 95,229
surface-surface, 209
definition, 47
Doppler broadening, 129, 131
transfer equations, 48
Effective line width, 135
Half width of line, 130
Effective band width, 142
Induced emission, 19, 121
Efficiency factor, 284
Intensity, 8
Einstein coefficients, 122
definition in medium, 9, 57
Elsasser model, 143
invariance along path in vacuum, 11
Emission, 4 , 2 1
Ionization potential, 6, 117
from volume, 23
Jump boundary condition, 72, 79, 82, 86,
induced, 19,121,122
line. 134
medium with nonunity refractive index, Kirchhoff's law, 26
315 Krakatoa, 295
spontaneous, 19,122 Langley, Samuel P., 1
Emission approximation, 62,67 Laser, 33
Emission coefficient, 23 Line,
Emittance of gases, broadening, 128
carbon dioxide, 5,149,184 shape parameter, 129
definition, 25,145 strong, 137
water vapor, 149,185 weak, 136
Energy conservation, 43,166,246 width, 130, 132, 133, 135
Energy density, 56 Line reversal technique, 342
Energy levels, 5,128 Local thermodynamic equilibrium, 32
Enclosure theory, 159 Lorentz profile, 131
band equations, 189 Luminous flames, 331
matrix of equations, 164 Luminous particles, 342
Equation of transfer, 37,38,125 Luminescence, 343
approximations to, table, 62 Mean absorption coefficient,
differential form, 40 incident, 46, 93
integral form, 41 Planck, 45,66,67,93,224
plane layer, 46,49 Rosseland, 78,93
photon model, 56, 125 Mean beam length, 175
Equilibrium gas not optically thin, 176, 179
local thermodynamic, 32 geomteric, 191, 192, 194
radiative, 45 optically thin limit, 178
Exchange factor approximation, 211, 218, table of values, 181, 192
260 Methane band radiation, 149

Mie scattering, 278, 299 Scattering from particles,

Milne-Eddington approximation, 100 diffraction from sphere, 292
Monte Carlo, 221, 223 large dielectric sphere with refractive
adjacent gray regions, 233 index near unity, 292
concentric cylinders, 231 large diffuse sphere, 289
parallel plates, 225, 229 large specular sphere, 286
Mie scattering, 278, 299
Natural broadening, 129, 131
Rayleigh scattering, 294
Net radiation method, 159
Scattering optical thickness, 304
Nonluminous flames, 329
Schuster-Schwarzschild approximation, 97
Optical thickness, 15, 40, 47, 3W,307
Schrodinger wave equation, 117
Optically thin limit, 53,67,178
Seeding with particles, 342
Oscillator strength, 127
Slip, 53, 72
Parallel plates, 166, 196, 199
coefficient, 255
black with gray gas between, 49
diffusion solution, 83
absorption coefficient, 333
Monte Carlo solution, 225,229, 231
concentration, 340
radiation and conduction, 247 electron~agnetictheory predictions, 335
Particle radiation, 342 optical properties, 336
Penetration distance, 14 total emittance, 338
Phase function for scattering, 284 Source function, 307
Photon, 6, 56 Spheres,
momentum, 117 concentric, 89, 109
Planck distribution, 121 Spontaneous emission, 19,122
Planck mean absorption coefficient, 45, Stark broadening, 129, 134
66,67,93,224 Statistical weight, 121
Polari~abilit~, Stimulated emission, 19
definition, 296 Strong line, 137
table for types of scattering, 296 Theoretical flame temperature, 325
Population inversion, 34 Thermodynamic equilibrium, local, 32
Radiative equilibrium, 45 Transfer equation, 38,40, 125, 302,306
Rayleigh scattering, 294 Transient problems, 237,246,345
cross section, 295 Transmittance, 29
phase function, 298 Transmittance factors,
Refractive index, geometric mean, 164, 168
nonunity, 315 geometrical, 164
Rosseland diffusion equation, 74, 75, 78 Transparent gas approximation, 62,64
Rosseland mean absorption coefficient, True absorption coefficient, 19,40,127
78,93 Uniform gas, 15, 159,205
Rydberg constant, 117 Vibration-rotation band, 7
Scattering, 238, 277 Water vapor radiation,
anisotropic, 20 band radiation, 149
elastic, 20, 277 charts, 185, 186
equation of transfer, 302, 306 mixture with CO?, 186
inelastic, 7, 20, 277 Wave function, 118
isotropic, 20 Weak line, 136
Scattering coefficient, 13, 20, 283 Wien distribution, 125
Scattering cross section, 280 Zoning method, 207



POSTMASTER: If Undeliverabfe (Section 158

Postal M n n d ) Do Not Return

"The aeronautical ar~dspace activities of the United State1 s b d

6e conducted SO as 80 contribute . to the expansion of humm
ktzowledge of phenomena in the at?l~~Iphere d space. The
Admini~trationshall provide for the widest practicable and
appropriate dissemination oj iufortnation concerning its activitiej
and the results thereof."


technical information considered Information published in a foreign
imponant, complete, and a lasting language considered to merit NASA
contribution to existing knowledge. disyibution in English.


broad in scope but nevertheless of Information derived from or of value
importance as a contribution to existing to NASA activities. Publications indude
knowledge. conference proceedings, monographs,
data compilations, handbooks, source-
TECHNICAL MEMORANDUMS: books, and special bibliographies.
Information receiving limited distribu-
tion because of preliminary data. TECHNOLOGY UTILIZATION
secprity classification, or other reasons. PUBLICATIONS: Information on tech-
nology used by NASA that may be
CONTRACTOR REPORTS: Scientific of particular interest in commercial and
and technical information generated other non-aerospace applications.
under a NASA contract or grant and Publications indude Tech Briefs,
considered an important contribution to Technology Utilization Reports and
existing knowledge. Notes, and Technology Surveys.

Details on the availability of these publications may be obtained horn:

Washington, D.C. 90546